Measuring value for
money in healthcare:
concepts and tools
Peter C Smith
Centre for Health Economics
University of York
September 2009
QQUIP and the Quality Enhancing Interventions project
QQUIP (Quest for Quality and Improved Performance) is a five-year research initiative of the Health Foundation. QQUIP provides independent reports on a wide range of data about the quality of healthcare in the
UK. It draws on the international evidence base to produce information on where healthcare resources are
currently being spent, whether they provide value for money and how interventions in the UK and around the
world have been used to improve healthcare quality.
The Quality Enhancing Interventions component of the QQUIP initiative provides a series of structured
evidence-based reviews of the effectiveness of a wide range of interventions designed to improve the quality
of healthcare. The six main categories of Quality Enhancing Interventions for which evidence will be reviewed
are shown below.
For more information visit www.health.org.uk/qquip
Acknowledgements
This work was funded by the Health Foundation under the QQUIP initiative. I should like to thank Sheila
Leatherman, Vin McLoughlin, Dale Webb and four splendid referees for their helpful comments.
Published by:
The Health Foundation
90 Long Acre
London WC2E 9RA
Telephone: 020 7257 8000
Facsimile: 020 7257 8001
www.health.org.uk
Registered charity number 286967
Registered company number 1714937
First published 2009
ISBN 978-1-906461-11-9
Copyright The Health Foundation
All rights reserved, including the right of reproduction in whole or in part in any form.
Every effort has been made to obtain permission from copyright holders to reproduce material.
The publishers would be pleased to rectify any errors or omissions bought to their attention.
Contents
Summary
6
1. Introduction
1.1. Why is VfM important?
1.2. What is value for money?
1.2.1.Allocative efficiency: guiding purchasing decisions
1.2.2.Technical efficiency: operational performance assessment
1.3. What is the unit of analysis?
1.4. Comprehensive or partial VfM measures?
1.5 Some examples of VfM performance measures
10
11
12
15
19
20
21
22
2. What are the components of VfM?
2.1. What is valued?
2.1.1. Health gain
2.1.2.The patient experience
2.1.3.Inequalities
2.1.4.Externalities and broader economic outcomes
2.1.5.Outputs: counting activity and processes
2.2.Valuing system outputs
2.3.What are the inputs?
2.4.Environmental constraints
2.5.Short run or long run?
29
29
29
32
33
33
34
35
36
37
39
3. Measuring VfM
3.1. Adjusting for environmental constraints
3.2.Analytic models of VfM
3.2.1.Statistical methods
3.2.2.Descriptive methods
41
41
43
44
45
4. Conclusions
47
References
51
Measuring value for money in healthcare: concepts and tools
Summary
Summary
Section 1
The concept of value for money (VfM) has been central to health policy and the delivery of healthcare
for some time. In its abstract form, the concept of VfM is straightforward: it represents the ratio of some
measure of valued health system outputs to the associated expenditure, and few would argue that its
pursuit is not a worthy goal. The main reasons for an interest in VfM relate to accountability: to reassure
payers, in particular taxpayers, that their money is being spent wisely, and to reassure patients that their
claims on the health system are being treated fairly and consistently.
1.1 In practice, the measurement of VfM is challenging and gives rise to some important
methodological questions. The main aim is to offer an understanding of how resources are
successfully transformed into valued health system outputs. But there are several stages to that
transformation, each of which can be measured with different degrees of accuracy and ease. The
result has been a profusion of partial indicators of VfM, but a relative dearth of definitive measures
that capture the whole transformation process in the form of a cost-effectiveness measure.
1.2 The two fundamental economic concepts underlying VfM are allocative efficiency and technical
efficiency.
1.2.1 Allocative efficiency indicates the extent to which limited funds are directed towards
purchasing the correct mix of health services in line with the preferences of payers. It is
central to the work of the National Institute for Health and Clinical Excellence (Nice),
which uses expected gains in quality-adjusted life years (QALYs) as the central measure
of the benefits of a treatment, and cost per QALY as a prime cost-effectiveness criterion
for whether or not to mandate adoption of a treatment by the NHS. The assumption
underlying this approach is that the taxpaying public wishes to see the taxes assigned to
the NHS used to maximise health gain. Thus, in deciding what services to purchase, the
main (but not sole) focus of allocative efficiency is prospective.
1.2.2 Technical efficiency is quite distinct. It indicates the extent to which a provider is securing
the minimum cost for the maximum quality in delivering its agreed outputs. The prime
interest in technical efficiency is in operational performance assessment and the extent
to which resources are being wasted. The main focus of technical efficiency is therefore
retrospective. This paper focuses mainly on retrospective VfM measurement.
1.3 In undertaking any VfM analysis, it is essential first to decide on the nature of the entity under
scrutiny. At one extreme this might be the whole health system. At the other extreme, it might
be the treatment of an individual patient.
1.4 Another fundamental decision is whether to seek out a comprehensive measure of the costeffectiveness of the entire entity or to rely on partial indicators of some aspects of VfM.
In the latter case, incompleteness can take two forms: omission of some aspects of the
transformation from resources to valued outcomes (for example, no health outcome data),
or omission of some of some of the functions of the entity (for example, analysis of only the
inpatient activities of a hospital).
1.5 There have been numerous efforts to implement VfM measurement schemes. These include
whole-system productivity estimates, as attempted by the World Health Organization (WHO)
in the World health report (WHR) 2000 and by the Office for National Statistics (ONS) in UK
trends over time. These comprehensive, whole-system measures are experimental. More
practical approaches have offered useful but incomplete indicators of VfM. All efforts have
encountered severe methodological challenges and lack of data in key domains.
6
Smith
Summary
Measuring value for money in healthcare: concepts and tools
Section 2
There are a number of components of VfM that need to be considered when developing any VfM
measure. These may include the eventual outcomes of interest, intermediate outputs and activities,
inputs, possible external constraints on achieving VfM, and whether a long or short time horizon is being
adopted.
2.1 Outcomes are the valued outputs of the health services. Although there is room for debate
about what is valued, in the NHS they can be grouped according to four broad categories:
health gains, the patient experience, inequalities, and the broader social and economic benefits
of health services.
2.1.1 Considerable progress has been made in the conceptualisation and measurement of
health gain in the form of QALYs and associated measurement instruments such as
EQ5D. These have been used mainly for the purposes of health technology assessment,
not for the routine surveillance of provider performance, and there are major challenges
involved in measuring health gain throughout the health system. However, the NHS
has recently made a start by mandating collection of ‘before and after’ health status
measurement for four common surgical procedures.
2.1.2 There is a growing acknowledgement that patients and their families place considerable
value on the experience of their interactions with the health services, independent of
health outcome. WHO grouped these patient experience concerns in the category
of ‘responsiveness’, which includes concepts such as choice, communication,
confidentiality, quality of amenities and prompt attention. Surveys of the patient
experience have become commonplace, and the challenge is to develop adequate
summary measures of provider performance.
2.1.3 Inequalities in health and inequalities in access to health services have been a persistent
cause for concern in many health systems. There are two broad schools of thought
on how to handle equity issues in VfM measurement. One looks to develop separate
measures of equity, based on divergence of outcomes for different social groups, for
example, while the other weights the outcomes of health system performance, such
as health gain, more heavily for disadvantaged groups. Although the latter approach is
probably more promising, quantification of differential weights is in its infancy.
2.1.4 Health services yield benefits beyond the immediate heath gain to patients, such as
increased worker productivity, increased personal independence and reduced burden
on carers and social care agencies. Depending on the context, there may be a case for
integrating these considerations into any VfM analysis, although measurement issues are
often challenging.
2.1.5 Because of data limitations, many VfM analyses are forced to rely on measures of
outputs (quantities of activities) rather than measures of the eventual outcomes for
patients and society. This can be unproblematic if the outputs are known to lead to
good eventual outcomes and there is known to be little variation in quality of providers.
However, it clearly can be seriously misleading if this is not the case.
2.2 Once the valued outputs have been identified, there will often be a need to combine them into
a single measure of attainment. Some exploratory research has explored the relative weights
people attach to diverse outcomes, such as waiting time, travel distance and health outcome,
using methods such as ‘stated preference’ experiments. However, this work is in its infancy,
and the analysis will usually have to rely on a rudimentary rule of thumb to combine different
outcome measures.
2.3 Inputs represent the ‘money’ component of the VfM analysis. They can be readily identified
if the units are discrete organisations such as hospitals. However, they can be much more
Smith
7
Measuring value for money in healthcare: concepts and tools
difficult to identify if the unit of analysis is smaller, such as a hospital department, as it
becomes increasingly difficult to estimate what fraction of the hospital’s resources are devoted
to producing the outputs of the department. There are a number of unresolved challenges
associated with costing methodology.
2.4 Different health service organisations work in the context of different external constraints, such
as the health characteristics of the local population, local transport, geography and economic
conditions, and the activities of other agencies both inside and outside the health sector. Any
comparative VfM analysis should take account of these differences.
2.5 The naive VfM assumes that contemporary inputs give rise to contemporary outcomes. Yet in
most healthcare there is a need to adopt a longer time perspective. Some of today’s outcomes
arise from health service endeavours, such as disease prevention, in previous periods. And
some of today’s endeavours affect outcomes only at some time in the future. Therefore, when
analysing the VfM of some services, it will be necessary to adopt a longer time horizon.
Section 3
Traditionally, efforts to measure VfM have been piecemeal and partial. Indicators such as inpatient
length of stay are helpful and suggestive, but tell only part of the VfM story, and can lead to inappropriate
responses and adverse consequences if not used with care. Technical analytic efforts have therefore
been directed at developing more comprehensive VfM measures to complement the partial indicators.
This section summarises some of that work.
3.1 A great deal of analytic effort has gone into developing methods of adjusting for the
environmental differences discussed in 2.4. The simplest approach is to compare only like with
like, by selecting for comparison only organisations working in similar environments and using
methods such as cluster analysis. However, this is a crude expedient, and researchers have
developed more subtle methods of risk adjustment to address some aspects of environmental
variation. These enjoy wide acceptance in some domains (such as adjusting surgical outcomes
for casemix), but are not so advanced in many other health services.
3.2 Two broad approaches have been adopted for developing ‘single number’ measures of an
organisation’s VfM. I class these as statistical and descriptive methods.
3.2.1 Statistical methods are based on the conventional econometric regression models.
A statistical model of costs seeks to estimate an organisation’s expected costs given
the outputs it produces. In its simplest form, efficiency is simply indicated by the
organisation’s observed deviation from this prediction. Approaches such as stochastic
frontier analysis (SFA) seek to decompose the deviation into a random element (not
caused by inefficiency) and an inefficiency element, the issue of interest. However, SFA
methods do not enjoy universal endorsement from researchers.
3.2.2 Descriptive approaches are based on the class of technique known as data envelopment
analysis (DEA). DEA searches for the organisations that ‘envelop’ all other organisations
on the basis of a composite estimate of VfM. For each organisation, it looks for all other
organisations that secure the same, or better, outputs with the lowest use of inputs.
Or, conversely, it can be used to search for the other organisations that use the same,
or lower, inputs to secure the highest level of outputs. For each organisation, the ratio
of actual to optimal performance is referred to as ‘inefficiency’. DEA can yield useful
information, but needs to be used with extreme care for the purpose of benchmarking
performance.
8
Smith
Summary
Measuring value for money in healthcare: concepts and tools
4. Conclusions
The report highlights two fundamental roles for VfM measurement: prospective assessment of
technologies (for resource allocation purposes) and retrospective assessment of the VfM of individual
providers (performance assessment). In combination, these roles comprise a major element of the
functions of healthcare purchasers (or PCT commissioners as they have become known in England).
The purchasing function is immensely complex and has hitherto been undertaken with only limited
success in most health systems (Figueras, Robinson et al, 2005). Concerted attention to VfM
measurement offers a central focus for improving purchasing for health and healthcare.
The resource allocation role of VfM measurement is relatively well understood, albeit mainly in the
context of individual treatments. By contrast, the performance assessment role of VfM measurement
is underdeveloped. Until now, there has been a reliance on partial indicators of VfM. These can act as
useful diagnostic tools, but can also give misleading signals if used carelessly. There is an urgent need
to complement these partial measures with more comprehensive measures of VfM performance. The
arguments in favour of pursuing increased comprehensiveness are:
• It offers a rounded assessment of an organisation’s performance across all domains of
endeavour.
• It can facilitate a focus on patient outcomes, regardless of the specific treatments or diseases
under consideration.
• It facilitates communication with ordinary citizens and promotes accountability.
• It indicates which of the entities under scrutiny represent the beacons of best VfM.
• It indicates which entities should be priorities for improvement efforts.
• It can stimulate the search for better data and better analytic efforts across all healthcare.
• It offers mangers of local organisations the freedom to set their own priorities and to seek out
improvements along dimensions of performance where gains are most readily secured, and it
does not seek to micromanage (a possible consequence of piecemeal VfM indicators).
Nevertheless, partial indicators also offer benefits:
• They can identify serious failings in some parts of the organisation, even if more aggregate
measures of VfM indicate no cause for concern.
• They offer a diagnostic tool for identifying what to attribute poor performance to, and therefore
what remedial action to take.
• They may be the only realistic approach if an attempt to be comprehensive leads to a reliance
on very feeble or opaque data in some dimensions of performance.
• When aggregating different dimensions of performance, comprehensive measures may have to
rely on preference weights that are highly contested.
The paper indicates that there are many challenges in embedding VfM considerations into the scrutiny
and improvement of the health system. It concludes by summarising the priorities for three key
constituencies: policy makers and regulators, managers, and researchers.
Smith
9
Measuring value for money in healthcare: concepts and tools
1. Introduction
1. Introduction
The pursuit of value for money has become the holy grail of health systems worldwide. It appears selfevident that policy makers should wish to deploy health system expenditure with the aim of securing
maximum value in the form of benefits to patients and the broader population. It may therefore seem
somewhat surprising that the topic of VfM generates fierce debate and controversy around methodology.
This paper describes the various concepts of VfM in common use, examines how VfM measures are
constructed, discusses the challenges inherent in measuring VfM, and assesses the priorities for future
efforts in this domain.
The concept of value for money is straightforward: it represents the ratio of some measure of valued
health system outputs to the associated expenditure, and few would argue that its pursuit is not a worthy
goal. But, in practice, discussion of VfM gives rise to some fundamental questions. What is valued? Can
we necessarily identify the volume of ‘money’ going into the health system? And what precisely is the
health system entity under scrutiny? It turns out that it is important to secure some clarity about these
and related issues if the notion of VfM is to be made operationally useful in guiding policy makers and
practitioners towards a health system in line with policy objectives.
It is worth noting at the outset that not all stakeholders necessarily advocate the pursuit of VfM. In
systems such as the English NHS, for example, in which patients do not personally bear the full costs of
their treatment, most patients are more interested in the effectiveness of healthcare rather than its VfM.
Indeed, they might well view the pursuit of VfM as antipathetic to their own interests in securing the best
possible treatment regardless of expense. Similarly, clinicians may view an interest in VfM with some
scepticism because it might place limitations on the healthcare they are able to offer, perhaps inhibiting
them from doing what they feel is best for their patients.
Yet, whatever the chosen system of finance, someone must always pay for the healthcare delivered, and
those payers’ interests are served by focusing on VfM. In particular, in a tax-funded system such as the
NHS, taxpayers want to be assured that their payments are being used in line with their objectives. It is
these objectives that create the concept of value in VfM, and the pursuit of VfM fundamentally reflects
a desire to respect the interests of payers (or their representatives). In England these might include the
Department of Health, primary care trusts (PCTs) and general practitioner commissioners, on behalf of
the principal funder, the taxpayer.
This payer perspective means that, at times, VfM considerations may come into conflict with movements
such as the safety agenda and ‘patient centredness’. The principle of VfM suggests that these agendas
should be pursued, but only up to a point, to the extent that they promote value for money. In short, from
a payer perspective, considerations of health system effectiveness are trumped by the notion of costeffectiveness, which embraces most aspects of VfM.
In writing this paper I have found it very difficult to present the ideas underlying VfM in a succinct and
transparent fashion. This may be due to my shortcomings as an author. However, I suspect that it also
reflects the complexity of the concept of value for money. It is, for example, noteworthy that of all the
Public Service Agreements (PSAs) developed by the UK government for the English public services, it
is those relating to VfM that have given rise to the most conceptual and operational challenges (Smith,
2007a).
I hope, nevertheless, that this paper offers some help in navigating the conceptual and analytic jungle
of VfM measurement. Throughout this paper, I concentrate on healthcare, and do not consider broader
10
Smith
1. Introduction
Measuring value for money in healthcare: concepts and tools
issues of health promotion and preventative care. The document is written for a UK readership;
however, the issues discussed are universal and should be relevant for most health systems. Section 1.1
discusses why VfM measurement is important, while section 1.2 examines what is meant by the concept
of VfM. Section 1.3 explains the need for a clear understanding of the entity under scrutiny in VfM
analysis. Section 1.4 discusses the tension between using comprehensive, but perhaps unattainable,
measures of VfM and the use of partial VfM indicators, which are practical, but possibly misleading and
incomplete. Section 1.5 gives some operational examples.
1.1. Why is VfM important?
VfM is, in practice, a subtle and multidimensional concept, and in many domains it is challenging to
develop satisfactory measures of VfM that are not misleading. The question therefore arises whether it is
sensible to seek to develop such measures. The answer must be a resounding ‘yes’, for two fundamental
reasons relating to accountability: to reassure payers that their money is being spent wisely and in line
with their intentions; and to reassure patients that their claims on health system resources are being
treated consistently and fairly. Pursuit of these objectives makes the search for good indicators of VfM
imperative. The alternative is to leave decision makers facing a cacophony of competing claims for
healthcare resources with no coherent methodology for reconciling those claims. Properly used, VfM
offers the only unifying concept with which to evaluate healthcare technologies, inform the allocation
of resources within the health system and the broader economy, and assess the performance of
components of the health system.
Furthermore, in most health systems, an increasingly rich information base is developing, offering
insights into many aspects of the epidemiology, costs, processes and outcomes of healthcare. The
Health Foundation’s Quest for Quality and Improved Performance (QQUIP) initiative has demonstrated
the enormous scope of data now available in the UK. It was set up to help answer three fundamental
questions about healthcare in England:
• What is the current state of quality and performance?
• What works to improve quality and performance?
• Are we getting value for money from what is spent on the NHS?
The QQUIP website (www.health.org.uk/qquip/) brings together data from a wide range of sources to
reveal national and international trends relating to diseases and quality of care, with over 150 charts on
priority areas such as cancer, heart disease, diabetes and mental health.
Data such as these are of immense importance in their own right. However, their value can be enhanced
further by scrutinising them within the overarching framework of VfM analysis. This can go beyond
the piecemeal scrutiny of individual data items to offer decision makers an evaluative framework. For
example, for a specific treatment, a comprehensive VfM analysis can combine separate information on
trends in costs, casemix, volume and outcomes in order to track the cost-effectiveness of the treatment
over time.
In assessing the role of VfM information in healthcare, it is usual to focus on its importance in promoting
the accountability of various ‘agents’ to their ‘principals’. Figure 1 illustrates just some of the numerous
agency relationships that exist within healthcare. The most obvious agency relationship is that between
clinician and patient. Flows of information are essential to this relationship, for example in optimising
patient care and informing patient choice. However, VfM concerns play relatively little part in the
clinician/patient relationship in universal health insurance systems such as the NHS, because the patient
has no direct interest in the clinician’s remuneration.
Smith
11
Measuring value for money in healthcare: concepts and tools
Summary
Figure 1: Some of the accountability relationships in healthcare
Clinician
Patient
Professional
Provider
organisation
Government
Purchaser
organisation
Citizens
By contrast, in almost all of the other agency relationships in healthcare, VfM plays a major role, because
the principal is a payer of some sort. For example, taxpayers want assurance that tax contributions to
the NHS are being used to best effect. Similarly, the government needs to know that the health system is
receiving the right amount of finance relative to other parts of the economy, and is using it appropriately.1
It therefore also requires assurance that the money it distributes to local purchasing organisations such
as PCTs is well spent. In turn, PCTs need to know that their funds are being spent in the best way, in
terms of the treatments they purchase and the organisations from which they commission care.
Notwithstanding the general concern with VfM, the different types of accountability relationship
lead to the need for different types of VfM information. NHS purchasers may require very detailed
benchmarking information in order to design and monitor contracts with specific providers. By contrast,
taxpayers may need quite aggregate and broad-brush information on productivity trends with which to
hold their government to account. There are also a number of components of VfM that address different
managerial concerns and require different approaches to measurement. These are considered in the
following section.
1.2. What is value for money?
The underlying intention in any VfM analysis is to offer insight into how resources are successfully
transformed into valued outcomes. There are a number of stages in this transformation (the care
pathway), and much of the confusion in discussing VfM arises because commentators discuss different
parts of that process. In principle, the VfM part of the transformation can be captured in the notion of
cost-effectiveness. However, the data demands of a full cost-effectiveness analysis are often prohibitive.
And, in any case, decision makers may often require more detailed diagnostic VfM indicators for just part
of the care process.
As an example, figure 2 illustrates a typical, though simplified, process associated with the treatment of
hospital patients. The overarching concern is with cost-effectiveness – the transformation of costs (on
1 Of course, this will require the development of analogous VfM measures in other sectors of the economy.
12
Smith
1. Introduction
Measuring value for money in healthcare: concepts and tools
the left-hand side) into outcomes (the right-hand side). However, there are a number of discrete stages in
that transformation. In the first instance, the money available to the hospital is used to purchase physical
inputs to the care process, for example in the form of labour, capital and drugs. These are used to
produce a range of activities that, in aggregate, create physical outputs in the form of episodes of patient
care. The outputs are not an end in themselves; rather, they offer certain qualitative characteristics that
are valued by patients. These are designed to create a desired outcome, centrally but not exclusively
related to improvements in the length and quality of life.
Figure 2: The production process in hospital care
Costs
Inputs
Activies
Outputs
Outcomes
Capital;
labour;
drugs
Diagnostic
tests;
surgical
procedure
Episode of
hospital care
Improvement
in length and
quality of life
Patient
experience
Most concepts of VfM refer to different aspects of the care process such as this, often offering a partial
insight into some aspect of the transformation process. For example, the familiar length of inpatient
stay metric offers an insight into the relationship between an output (an episode of hospital care) and
one specific input. This is partial in two senses: it does not embrace all of the transformation process
(ignoring outcomes and costs), and it does not include all the resources used. Nevertheless, it often
offers a useful insight into some aspects of hospital VfM.
In general, the VfM terminology is confusing and ambiguous. It is therefore not always clear which
aspect of the transformation process a commentator is referring to. Economists and accountants use
slightly different concepts; in particular, the notion of efficiency can be used to refer to various aspects of
the production process.
Take the transformation of money into inputs. There are two aspects of this process. First, are
the inputs purchased at minimum cost (sometimes referred to as ‘economy’)? For example, is the
organisation paying wages that are higher than the local market rates? And second, has the correct
mix of inputs been put in place? For example, is the organisation employing the right mix of doctors,
other professionals and administrators, thereby avoiding the wasteful use of skilled personnel on routine
tasks?
The production process now moves to the creation of physical outputs, usually in a hospital setting
and referring to single episodes of patient care. There is considerable scope for waste in this process,
for example in the form of duplicated or unnecessary diagnostic tests, the use of branded rather than
generic medicines, or unnecessarily long stays. Much depends on how the internal processes of the
hospital are organised in order to maximise outputs for given inputs. Accountants refer to the success of
converting physical inputs into outputs as ‘efficiency’. Rather confusingly, economists refer to it under a
number of headings: technical efficiency, managerial efficiency, or even x-efficiency.
Technical efficiency makes no judgement on how much the outputs are valued by society. The next
concept to be addressed therefore is whether the hospital is producing the ‘right’ mix of outputs. For
example, a hospital may produce outputs (episodes of care) with great technical efficiency, but society
may value certain outputs (such as the treatment of glue ear) much less than other outputs that could be
Smith
13
Measuring value for money in healthcare: concepts and tools
1. Introduction
produced with the same resources. The extent to which the outputs of the organisation are maximised
in line with society’s valuation2 of their characteristics is measured using the concept of allocative
efficiency.
Finally, an important aspect of healthcare is that there is great scope for variation in effectiveness. This
is often referred to as the ‘quality’ of the outputs produced, arising from variations in clinical practice and
competence. The notion of quality in healthcare has a number of connotations. However, in this paper,
I use it to refer to two broad concepts: the clinical outcomes achieved (usually measured in terms of the
gain in the length and quality of life), and the patient experience (a multidimensional concept, discussed
further in section 2). So, for example, even though two hospitals may produce identical numbers of hip
replacements, owing to variations in clinical practice and competence, the value they confer on patients
(in the form of length and quality of life, and the patient experience) may vary considerably. Qualityadjusted output is usually referred to as the ‘outcome’ of care in the productivity literature. The quality
of care has become a central concern of policy makers, and its measurement is usually essential if a
comprehensive picture of VfM is to be secured.
In summary, VfM can be examined in a number of ways, including:
• the economy with which physical inputs are purchased
• the extent to which the chosen inputs are combined in an optimal mix
• the technical efficiency with which physical inputs are converted into physical outputs
• the allocative efficiency of the system’s chosen outputs
• the quality of the care provided (its effectiveness).
Each of these concepts scrutinises a particular aspect of the transformation process. However, the holy
grail of value for money is the notion of cost-effectiveness, the ratio of eventual outcomes to the costs
incurred, which embraces the entire production process and therefore all the separate VfM concepts
mentioned above. All the other measures give important diagnostic information because they allow us to
pinpoint where inefficiencies are arising. However, it is important to recognise that they give only partial
insights into healthcare VfM.
The ideas of allocative and technical efficiency can be illustrated diagrammatically (see figure 3).
Suppose there are just two outputs (treatments perhaps). Given current technology, for a given level of
expenditure, the maximum feasible production of the two outputs is represented by the curve FF; that
is, all organisations must lie on or below FF. Technically efficient organisations can lie anywhere on the
frontier FF. This traces the maximum attainable outputs of the organisation given its current budget. The
curvature reflects the increasing difficulty of squeezing out an additional unit of either output. However,
the frontier makes no judgement on the relative value of the two outputs. The relative value to society of
the two outputs can be represented by the slope of a line such as CC, which indicates the value of output
1 relative to output 2.3 Only one point on the frontier P* then is allocatively efficient. It is where the sum of
the outputs, weighted by society’s values, is maximised. Anywhere else on the frontier FF produces less
societal value.
Of course healthcare organisations produce many more than two types of output. Furthermore, to make
this principle operational, we should require information on the value of all the different services, which,
in practice, is a large gap in our current knowledge (see section 1.2.1). However, the principle remains
2 I discuss in more detail how these valuations might be derived in section 2.
3 The assumption here is that the relative valuations remain constant across all levels of production. This need not always be so, in
which case CC also becomes non-linear. However, the argument does not change.
14
Smith
1. Introduction
Measuring value for money in healthcare: concepts and tools
Figure 3: Allocative and technical efficiency with just two outputs
Output 2 c
F
•C
• P*
•A
•B
F
c
Output 1
unchanged: organisations should, in theory, seek to maximise aggregate value according to a societal
judgement on the relative values of the individual services they provide.
Only organisations located at P* in figure 3 secure full technical and allocative efficiency. Some
organisations such as A might produce the right mix of the two outputs but they lie within the frontier, so,
although allocatively efficient, they are technically inefficient – they could produce more of each output.
Other organisations, such as B, might lie on the optimal frontier, but produce an inappropriate mix of
outputs, so are allocatively inefficient. In this case, B should produce more of output 2 at the expense of
output 1, given society’s valuations. Yet other organisations, such as C, neither produce the right mix of
outputs nor lie on the frontier, so are both technically and allocatively inefficient They have chosen the
wrong mix of outputs and they are not producing as much of them as they could. The next two sections
introduce the two ideas of allocative and technical efficiency in broad conceptual terms, relating them to
the two fundamental managerial tasks of purchasing decisions and performance assessment.
1.2.1. Allocative efficiency: guiding purchasing decisions
A fundamental requirement in all health systems is to determine where the limited funds available
for healthcare and health promotion are best spent. The principles of VfM suggest that the objective
should be to maximise the cost-effectiveness of the health system, but this begs the question of what
we mean by effectiveness. This is an issue discussed in more detail in section 2.1 below, but the
dominant assumption in much VfM analysis has been that the prime objective should be to pursue the
maximisation of the health gain generated by the health system.4 In principle, one could measure health
gain crudely by the years of life added through the intervention of the health system. However, this is
manifestly unsatisfactory as it ignores variations in the quality of life. The usual approach to measuring
such gain has therefore become the ‘quality-adjusted life year’ (QALY), or its disability-adjusted life year
(DALY) counterpart, as explained in box 1 (Gudex and Kind, 1988).5
4 Other possible objectives are discussed in section 2.1.
5 Similar arguments underlie other measures of health status, such as the diability-adjusted life year (DALY).
Smith
15
Measuring value for money in healthcare: concepts and tools
1. Introduction
Box 1: The quality-adjusted life year
The QALY is a year of life, adjusted for the quality (or value) of life it offers to the individual. A year
in perfect health is considered equal to 1.0 QALY. Conversely, death would be given a weight of
zero. The value of a year in ill health is adjusted depending on the severity of the condition. For
example, a year bedridden might be given a value equal to 0.5 of a QALY. The QALY values of
imperfect health should, in principle, reflect the extent to which individuals would be prepared
to exchange a year of healthy life for a year in the unhealthy state. Thus, in this example, the
value of 0.5 suggests that a period of six months of life in perfect health is equivalent to one year
bedridden. This equivalence means that the QALYs from different treatments can be directly
compared, even though the quality of life conferred by the treatments may be very different.6
In its simplest form, cost-effectiveness is the ratio of health gain to additional expenditure incurred.
Its application suggests that decision makers should create a league table of the expected costeffectiveness of individual health technologies, as defined by the cost of securing an additional qualityadjusted life year. Williams (1985) gave a celebrated early example of a cost-effectiveness league table,
summarised in table 1. Note that, for each specific condition, the table includes only the most costeffective technology available.
Table 1: An example of an incremental cost per QALY league table
Pacemaker for atrioventricular heart block
£700
Hip replacement
£750
Valve replacement for aortic stenosis
£900
CABG (severe angina; left main disease)
£1,040
CABG (severe angina; triple vessel disease)
£1,270
CABG (moderate angina; left main disease)
£1,330
CABG (severe angina; left main disease)
£2,280
CABG (moderate angina; triple vessel disease)
£2,400
CABG (mild angina; left main disease)
£2,520
Kidney transplantation (cadaver)
£3,000
CABG (moderate angina; double vessel disease)
£4,000
Heart transplantation
£5,000
CABG (mild angina; triple vessel disease)
£6,300
Haemodialysis at home
£11,000
CABG (mild angina; double vessel disease)
£12,600
Haemodialysis in hospital
£14,000
Source: Briggs and Gray (2000) adapted from Williams (1985)
6 Of course individuals may differ considerably in their valuations of different health states, and their views may change depending
on whether or not they are already in poor health. There is a continuing academic debate on how to handle this variability, but for
the purposes of this paper I have assumed that a legitimate authority would be able to resolve such conflicts and impose a single
set of preferences.
16
Smith
1. Introduction
Measuring value for money in healthcare: concepts and tools
In order to decide which procedures to include in the funded package of care, the decision maker must
move down the table, starting with the interventions with maximum cost-effectiveness (lowest cost per
QALY). Additional procedures are included in the package until the healthcare budget is exhausted.
Of course, the volume of expenditure consumed by each intervention will depend on the incidence of
the associated disease. The chosen list of interventions creates the ‘health basket’ with the maximum
feasible cost-effectiveness within the budget constraint. The marginal treatment (the last one to be
included in the basket) indicates the level of cost-effectiveness against which every new technology
should be assessed, often referred to as the cost-effectiveness threshold. Box 2 describes one of the
earliest efforts to put these principles into practice.
Box 2: The Oregon Health Plan
One of the earliest efforts to implement cost-effectiveness principles in selecting a health
basket was the celebrated Oregon initiative, under which an extensive deliberative process
was undertaken to infer societal values and rank the treatments to be included in the state’s
Medicaid healthcare package for low-income residents (Eddy, 1991). Research evidence and
professional opinions on the effectiveness of different treatments were used, together with the
values derived from the public consultations, to draw up a list of around 700 pairs of conditions
and treatments to be given priority for funding. The Oregon Health Plan was launched in 1994
with funds available from the legislature to provide 565 out of 696 treatments on the final priority
list. The treatments included the bulk of preventative and curative services, with high priority
being attached to palliative care as a result of values identified during the public consultations.
The principal exclusions were the treatment of self-limiting conditions and conditions where no
effective interventions were available.
Adapted from Ham (1998)
When a new technology emerges, it should be assessed in the light of this threshold. If its costeffectiveness is superior to the threshold (and to the cost-effectiveness of any existing treatment for
the disease in question), then it should be included in the basket. If the budget remains unchanged, this
implies that a certain volume of the most marginal existing treatments may have to be removed from
the basket to make way for the new treatment. The number of treatments squeezed out in this way will
depend on the volume of expenditure required to meet the spending needs of the new treatment. Equally,
of course, if an existing treatment is found to be not cost-effective, its removal from the basket may allow
a certain number of previously excluded treatments to be included.
This is a deliberately naive view of the resource allocation problem and must often be moderated by
considerations of equity and broader societal objectives. However, it forms the bedrock of the operation
of the National Institute for Health and Clinical Excellence (Nice) and analogous health technology
agencies. It should also, in principle, inform the design of clinical guidelines and broader health policy
strategies. NICE is increasingly seeking to integrate VfM into its general treatment guidelines (Wailoo,
Roberts et al, 2004).
The notion of cost-effectiveness should inform the operations of local purchasers of healthcare, such as
PCTs, as well as national policy making. Thus, for example, if Nice judges a new treatment to be highly
cost-effective, and that it should be universally adopted, individual PCTs must in principle scrutinise all
their existing treatments to determine which offers the least benefits in relation to costs and remove it
from the local health basket. This process should continue until expenditure on the new treatment can be
accommodated within the local budget. In practice, of course, such scrutiny is usually not feasible, often
because of limitations in data availability and local capacity to undertake the necessary analysis.
Smith
17
Measuring value for money in healthcare: concepts and tools
1. Introduction
Moreover, as purchasing moves to the local level, local capacity and operational constraints
become important considerations. For example, local providers may have slack capacity in certain
specialties that, at least in the short term, might reduce the opportunity cost of implementing some
new technologies relative to the assumed national costs. Conversely, in the short term, there may be
substantial investment costs involved in implementing certain new technologies, perhaps because they
involve major reconfiguration of services. Furthermore, there will usually be limited capacity to implement
a large number of service changes. National measures of VfM may therefore have to be modified in the
light of local circumstances.
Other local considerations may become important when interpreting national (or international)
guidance. For example, local entities may be unable to reap the scale of economies assumed in national
calculations; local populations may vary in demographic and epidemiological characteristics from the
national norm; and variations in local preferences may be important, especially in health systems that
seek to offer some local political autonomy in healthcare purchasing choices.
Analytic approaches such as programme budgeting and marginal analysis (PBMA) have been
developed to guide local decision makers in making allocative decisions according to VfM criteria when
there are significant local considerations (Mitton and Donaldson, 2004; Ruta, Mitton et al, 2005). For
example, in the short to medium term, local decision makers may be constrained by factors outside their
direct influence, such as the existing configuration of local hospitals, the distribution and interests of local
GPs, and the supply of local social services (Birch and Gafni, 2002). Furthermore, local management is
constrained in its capacity, and cannot practically address all necessary service changes immediately.
The role of PBMA is to understand these limitations and to identify a feasible way of setting priorities that
will lead to concrete gains in value for money.
Box 3: Stages in priority setting using programme budgeting and marginal analysis
Determine the aim and scope of the priority setting exercise: Will the analysis examine changes
in services within a given programme or between programmes?
Compile a programme budget: The resources and costs of programmes combined with activity
information
Form a marginal analysis advisory panel: The panel should include key stakeholders (managers,
clinicians, consumers, etc) in the priority setting process
Determine locally relevant decision making criteria: The advisory panel determines local
priorities (maximising benefits, improving access and equity, reducing waiting times, etc) with
reference to national, regional and local objectives
Identify where services could grow and where resources could be released through improved
efficiency or scaling back or stopping some services: The panel uses the programme budget
along with information on decision making objectives, evidence on benefits from service,
changes in local healthcare needs, and policy guidance to highlight options for investment and
disinvestment
Evaluate investments and disinvestments: Evaluate the costs and benefits for each option and
make recommendations for change
Validate results and reallocate resources: Re-examine and validate evidence and judgements
used in the process and reallocate resources according to cost–benefit ratios and other decision
making criteria
Source: Peacock, Ruta et al (2006)
18
Smith
1. Introduction
Measuring value for money in healthcare: concepts and tools
Box 3 summarises the PBMA process suggested by Peacock, Ruta et al (2006). It highlights the central
role of assessing the local situation and the importance of seeking out ways to move incrementally
towards a more cost-effective configuration of local services. Note that this marginal approach is the
antithesis of the classical economic evaluation of health technologies described earlier in this section,
which implicitly assumes that the only constraint is one of financial resources.
The principles of health technology assessment relate primarily to the prospective assessment of
whether healthcare providers should adopt new technologies, and this section has emphasised the
importance of allocative efficiency in prospectively guiding purchasing decisions. However, there is
often a case for also embedding allocative efficiency considerations in retrospective performance
assessment. It may, for example, be important to ensure that providers are providing services in line with
the purchaser’s intentions and not diverting resources to services that have low societal valuations. For
example, a PCT may purchase ophthalmology services assuming a certain treatment threshold for, say,
cataract surgery. If the purchaser relaxes that threshold, it breaches the purchaser’s allocative efficiency
assumption and therefore compromises VfM maximisation.
1.2.2. Technical efficiency: operational performance assessment
The allocation decision usually assumes that the chosen health basket will be delivered to maximum
effect: that is, it is assumed that a certain level of cost-effectiveness will be secured for each of the
treatments in the basket. There is a quite distinct concern about whether providers carry out their chosen
activities in line with this assumption. This aspect of VfM reflects concerns about technical efficiency
and, in contrast to the allocative perspective, usually adopts a retrospective, performance assessment
focus.
Technical efficiency should be a central concern of national regulators such as the Audit Commission
and the Care Quality Commission in England, which seek to determine, for example, whether unit
costs of individual providers are excessive. Partial indicators of technical efficiency, such as average
length of stay, abound, but the same overarching criterion of cost-effectiveness that underlies allocative
efficiency should inform the analysis of technical efficiency. It therefore seems natural to include
quality (effectiveness) issues, as well as quantity of outputs produced, within the ambit of technical
efficiency wherever feasible. This broader concept of technical efficiency moves the analysis closer to
a retrospective measure of cost-effectiveness. It seeks to determine whether specific providers have
produced the expected health benefits at the lowest feasible expected costs.
Independent engineering standards rarely indicate the maximum attainable level of technical efficiency.
The prime instruments for assessing technical efficiency have therefore become various benchmarking
tools, which seek to compare different providers using partial indicators of VfM such as length of stay
and unit costs, and usually focus on specific diseases or treatments. While these allow individual
organisations to focus on apparent examples of good and bad practice, they are usually piecemeal
and incomplete. Moreover, in interpreting input and output data, one must take account of variations
in the circumstances of the different entities under scrutiny, often in the form of variations in patient
characteristics or disease severity. This is essential if one is to gain insight into how much apparent
variation in VfM can be attributed to the health organisation. Attribution is often addressed using
techniques such as risk adjustment; these are discussed further in section 3.1.
VfM benchmarking was a central concern of the very earliest performance indicators distributed
to English health authorities in the 1980s. These early data contained a number of rudimentary
measures of unit costs, and were intended ‘to help [managers] to assess the efficiency of the services
for which they are responsible’ (Department of Health and Social Security, 1983). Once the internal
NHS market became established, such benchmarking initiatives fell out of favour, perhaps because it
was believed that market forces would naturally encourage local purchasers and providers to pursue
Smith
19
Measuring value for money in healthcare: concepts and tools
1. Introduction
efficiency. However, the development of a suite of VfM indicators by the NHS Institute for Innovation and
Improvement (described in section 1.5 below) suggests that there is an acknowledged need for VfM
benchmarking data even in a more market-oriented environment.
Recognising the limitations of piecemeal comparison, analysts have developed a range of statistical
and management science techniques. These seek to assess the global technical efficiency of individual
institutions, based on measures of total inputs and total outputs (or outcomes where quality is known).
Such measures attempt to measure the ratio of all outputs, aggregated in some fashion, to all inputs.
Examples include techniques such as data envelopment analysis and stochastic frontier analysis,
discussed further in section 3.2.
1.3. What is the unit of analysis?
Fundamental to any examination of VfM is the need for clarity about the nature of the entity under
scrutiny and the scope of the associated analysis. At the micro end of the spectrum, the entity might be a
single treatment, the intention being to assess its value (benefit) in relation to cost. The health technology
assessment movement has made enormous progress in developing methodologies with which to assess
the cost-effectiveness of individual treatments (Drummond, Sculpher et al, 2005). Some of the most
advanced methods embed costing methodologies within clinical trials, so that inputs and outcomes can
be directly aligned. Yet agencies such as Nice can examine only a fraction of the technologies used
in healthcare, and have mainly concentrated on recent technological innovations. There is often little
evidence concerning the VfM of established technologies, and the VfM of much healthcare remains an
article of faith rather than an established fact. In short, great strides have been made in the methodology
for examining the VfM of treatments. However, it has been satisfactorily applied to only a fraction of
healthcare, and putting in place the research capacity needed to provide a broader evidence base is a
daunting prospect.
At the macro end of the spectrum, the most challenging task is scrutiny of the VfM of the entire health
system, defined by WHO as ‘all the activities whose primary purpose is to promote, restore or maintain
health’ (World Health Organization, 2000). In practice, this definition has proved very difficult to make
operational. Health system outcomes are usually defined mainly in terms of the health of the population.
However, many variations in mortality and disability appear to be beyond the direct control of the health
system, as defined, and even identifying all the inputs that comprise the health system is challenging.
A more usual form of retrospective VfM study therefore seeks to identify the performance of meso-level
entities, such as specific practitioners, teams, hospitals or other organisations within the health system.
Here the challenge is that such organisational entities may be operating in quite different circumstances,
perhaps because the population being cared for or the patients being treated differ markedly. Usually,
some form of risk adjustment becomes essential if meaningful comparison is to be made.
Emerging data capacity, in the form of individual patient records and long-term household surveys, is
now making it increasingly feasible to focus on the individual as the basic unit of VfM analysis. Often this
will take the form of a patient’s episode of care, requiring estimates of the resource inputs devoted to
the patient and the outcomes secured in a circumscribed setting. The European HealthBASKET project
demonstrated how this could be done on the inputs side for ten patient vignettes (Busse, Schreyögg et
al, 2008). The project yielded estimates of variations in costs between individual patients, hospitals and
countries. However, it did not consider outcomes, and demonstrated the major challenges involved in
assigning resource use to individual patients.
Furthermore, it may on occasions be important to extend the analytic perspective to a whole population
basis to capture individuals who may have benefited from but have not received treatment. In a similar
20
Smith
1. Introduction
Measuring value for money in healthcare: concepts and tools
vein, for many patients with chronic conditions, it may be more appropriate to move beyond discrete
episodes of care and examine the cost-effectiveness of, say, the year of care provided by the health
system in whatever provider setting.
There is no simple answer to the question of what the appropriate unit of analysis might be. There are
considerable methodological challenges whichever unit is selected. However, as a general principle, it is
important that the analysis reflects an entity for which there is clear accountability, whether it is the whole
health system, the health services organisation or the individual patient. Only then can the relevant
agent, whether it is the government, management board or physician, be held to account for the level of
performance revealed by the analysis.
1.4. Comprehensive or partial VfM measures?
Whatever the unit of analysis, a major decision in VfM analyses is whether to attempt to develop a
comprehensive measure of VfM, embracing all the major inputs and outputs of the whole entity under
scrutiny, or to resort to partial indicators of VfM. The attraction of comprehensive measures is obvious,
and is the ideal pursued by NICE in its evaluation of treatments. Yet there is a powerful argument that
partial VfM measures also offer useful insights, especially when seeking to diagnose the reasons for
poor VfM. This section considers the various approaches, but it should be noted that, for many purposes,
it is helpful to have available both comprehensive and partial VfM metrics.
Table 2 illustrates the various types of completeness available for hospital comparisons. In the top lefthand cell, the analysis might assess all the health outcomes and all the costs associated with a hospital
to develop comparative measures of whole hospital cost-effectiveness. Although the methods described
in section 4 aspire to this ideal, data limitations make it very challenging to implement practically. A more
modest ambition might be to compare hospitals only on casemix-adjusted costs (top right-hand cell)
without reference to clinical quality. Alternatively, the comparison might seek to use the comprehensive
principle of cost-effectiveness as a basis for comparison, but only for a selected treatment (bottom lefthand cell). Finally, the most modest analysis offers an incomplete measure of VfM for only part of the
hospital’s activity (bottom right-hand cell).
Table 2: Varying levels of completeness in measuring VfM
Total VfM
Partial VfM
Whole entity
Whole hospital performance
assessment
Hospital reference cost index
Part of the entity
Cost-effectiveness measures for
individual treatments
Average length of inpatient stay for
selected treatment
Whatever the aspect of healthcare under scrutiny, in principle, comprehensive cost-effectiveness
measures of VfM should embrace all the relevant outcomes of healthcare, intended and unintended.
As discussed in more detail in section 2, such outcomes are often summarised under two broad
headings: health gain and the patient experience. However, in some circumstances, they can extend
to broader societal objectives, such as enhanced worker productivity or reduced demands on patients’
carers. The first challenge when constructing a comprehensive measure is therefore to enumerate
and measure the various outcomes of relevance. It is noteworthy that NICE methodology has until now
concentrated on health benefits, and incorporating broader benefits poses considerable methodological
challenges. However, there are growing demands to move in that direction (House of Commons Health
Select Committee, 2008).
Smith
21
Measuring value for money in healthcare: concepts and tools
1. Introduction
In developing a comprehensive measure, the associated measures of outcome must be combined
according to some measure of the relative value of each outcome. This aggregation is essential if
valid comparisons are to be made between different treatments for the same disease, and between
different treatments for different diseases. Only if the relative benefits of treatments can be assessed in
a common currency is it possible to make informed judgements about the comparative VfM of different
treatments and different organisations.
In some domains, there has been extensive research on estimating values – for example, methodologies
underlying the QALY have derived estimates of the trade-off between quality of life (in the form of pain,
mobility, etc) and length of life implicit in many treatments. Although there exist large interpersonal
variations, and therefore continued debates about how to infer a societal set of values, this approach
permits comparison of diverse treatments in a common currency, a fundamental requirement for
developing a comprehensive measure of outcome (as summarised in box 1). However, in areas beyond
health-related quality of life, valuation is at a rudimentary stage of development – for example, there is
little evidence on how much citizens are prepared to trade off, say, waiting time against the clinical quality
of care (see section 2.2).
Whatever methodology is employed, once a composite measure of outcome has been derived, it can
be compared with the inputs (expenditure) to derive an estimate of VfM. Even here, however, there are
challenges. It can be quite challenging to estimate the inputs associated with the entity under scrutiny. In
particular, in hospitals, many of the costs are associated with various forms of overheads, and it can be
difficult to attribute the inputs to a specific treatment, department or team (section 2.3).
Also, on the inputs side, it should be noted that a comprehensive VfM measure might also have to
include expenditure not directly borne by the health sector. For example, some treatments or delivery
methods might impose substantial private costs on patients and their carers that are not borne by the
health system. The issue of who pays or value for whose money has received little attention to date, but
might become increasingly important as NICE broadens its remit. Furthermore, as with partial indicators,
it is often important to make some sort of adjustment for the environment within which each of the
entities is operating, such as the complexity of casemix (see section 2.4).
Note that comprehensive measures of VfM should also, in principle, accommodate the longer time
perspective. Many of the inputs to patient care take place over a number of years – for example, in
the form of preventative care – so merely comparing current inputs with current outputs may give a
misleading picture of VfM. Comprehensive VfM measures may therefore have to embrace quite long
time horizons. This issue is discussed further in section 2.5.
1.5. Some examples of VfM performance measures
This section illustrates the principles set out above with some examples of performance assessment
efforts to date. At the most ambitious level, WHO sought to derive a comprehensive measure of health
system performance in its World health report (WHR) 2000, which derived estimates of the costeffectiveness of the health system in each of its 181 member countries (World Health Organization,
2000). Box 4 describes the variables used in deriving this VfM measure. Five outcomes were specified,
alongside one input of health system expenditure. In addition, an adjustment was made for the level of
national development (as measured by average years of schooling). The WHO exercise provoked a
vigorous debate in policy and academic circles, and the response highlighted the enormous challenges
involved in deriving whole-system VfM measures (Anand, Ammar et al, 2003). In short, it demonstrated
that there are major issues still to be addressed in conceptualising the notion of the health system, in
measuring and valuing health system outcomes, and in quantifying the contribution of the health system
to outcomes.
22
Smith
1. Introduction
Measuring value for money in healthcare: concepts and tools
Box 4: Variables used in World health report 2000 model of health system performance
Outcomes
1. Overall health outcomes (measured by disability-adjusted life expectancy)
2. Inequality in health (measured by an index based on child mortality)
3. Overall health system responsiveness, reflecting respect for persons and client orientation (as
assessed by a panel of 1,791 key informants)
4. Inequality in health system responsiveness (as assessed by the key informants)
5. Fairness of financing (based on the proportion of non-food expenditure spent on healthcare)
Inputs
6. Expenditure per capita (from National Health Accounts)
Adjusted for
Average years of schooling (as a proxy for level of national development)
An alternative approach, adopted by the UK Office for National Statistics (ONS), is to track the changes
in productivity of a single system (the NHS) over time. This work arose from the 2005 report by Sir Tony
Atkinson recommending changes in the way that public service outcomes are measured in the national
accounts (Atkinson, 2005). It has led to the publication of a series of articles reporting methodological
progress in this area, and has also contributed to methodological developments at the Organisation for
Economic Co-operation and Development (OECD) (Smith and Street, 2007).
The principles underlying the ONS work are that (with the notable exception of the preventative area, the
treatment of which remains underdeveloped) the outcomes of the NHS are the aggregation of hundreds
of different types of activity, such as a specific hospital treatment, a GP encounter or a nurse visit (Office
for National Statistics, 2004). In aggregating such activities, a crucial issue then is how to attach a value
to each of these activities. Hitherto this has been done using the costs of the various activities as their
weights. This expedient is practicable but patently misleading because, for example, some very costly
procedures might be yielding very low patient benefits. It is therefore acknowledged that adoption of
what are known as value weights, under which the activities are weighted according to the benefits they
confer on patients, is desirable (Castelli, Dawson et al, 2007). However, the limitations discussed above
in assessing the VfM of individual treatments preclude any such move in the foreseeable future (Smith
and Street, 2007).
Furthermore, the original ONS methods ignored issues of effectiveness, yet, as can be seen on the
QQUIP website, there is clear evidence that the quality of healthcare is changing over time. One way
to capture such changes is to incorporate measures of quality as adjustments to the counts of relevant
activities in the cost-weighted output index. An example of this approach has been developed by the
University of York and the National Institute of Economic and Social Research (Dawson, Gravelle et al,
2005). The English Department of Health and ONS have used some partial measures of quality (waiting
times for non-emergency inpatient care, 30-day survival after hospital admission for certain inpatient
Smith
23
Measuring value for money in healthcare: concepts and tools
1. Introduction
activities, blood pressure control for patients in general practice) to demonstrate the method (Office for
National Statistics, 2006).
The principle adopted has been to apply the selective quality measure to relevant activities before
aggregating into the output index. For example, trends in the numbers of certain inpatient episodes
are weighted by concurrent trends in post-operative survival rates to yield quality-adjusted time series
of outcomes before they are aggregated into the cost-weighted activity index. The present ONS
assumption is that there is a one-to-one relationship between such quality measures and the outcome of
interest. I shall argue below (section 2.1.1) that this argument is open to challenge.
Figure 4 illustrates the trends in productivity over a ten-year period as produced by the ONS
methodology under a variety of assumptions. The first four series use only a quantity measure of output
(NHS activities weighted by cost), while the last four seek to include quality measures, principally in the
form of post-operative survival rates. The results suggest a high level of sensitivity to the assumptions
used, with changes in productivity since 1999 ranging from an average annual fall of 1.3 per cent (series
1) to an average annual rise of 0.2 per cent (series 8). It is noteworthy that the incorporation of quality
adjustments markedly improves the estimates of productivity change, reflecting a steady improvement
in certain aspects of NHS quality since 1999. ONS has subsequently incorporated the results of
consultations on methodology, and more recent work focuses on a narrower range of options (Office for
National Statistics, 2008). Over the period 2000 to 2006, these results indicate a drop in productivity of
2.5 per cent per annum with no quality adjustment, and 2.0 per cent per annum if quality improvements
are taken into account.
As well as measuring VfM at the treatment level and the whole-system level, there have been numerous
attempts to develop measures of the VfM of all types of organisations and practitioners within the
health system in a huge range of settings (Hollingsworth, 2003). Such measures are based on the
analytic statistical models described further in section 3.2. They undoubtedly offer some insights into
organisational performance. However, they often treat the organisation as a ‘black box’, and do not
pinpoint where in the production process inefficiencies are arising. Furthermore, to satisfy the need
to be comprehensive, they often have to rely on very questionable data; most notably, they rarely use
adequate outcome data. These measures therefore usually stop short of being fully comprehensive, and
can be be unreliable and hard to interpret. As a result, analytic effort has concentrated on developing
partial measures of VfM, as a practical response to the difficulty of developing comprehensive measures
and to provide more operationally useful information about VfM.
Some of the earliest partial measures include the unit costs of a single aspect of treatment, such as an
episode of inpatient care. These offer a summary measure of cost efficiency and indicate the extent
to which a) inputs are being purchased at minimum price, b) the organisation is deploying them in an
optimal fashion (that is, in the correct allocative mix) and c) the organisation is operating them with
optimal technical efficiency. Valid comparison of unit costs requires the units of physical output to be
comparable (they must entail treatment of identical types of patient) and the quality of outcome to be
identical. If these conditions do not hold, then proper VfM comparison requires an extended analysis to
embrace variations in outcome measures.
In practice, hospitals treat an extraordinarily heterogeneous mix of patients, so the most rudimentary
requirement for valid comparison is to adjust for variations in casemix between hospitals. The celebrated
system of diagnosis-related groups (DRGs) was originally developed with such adjustment in mind, so
that the actual costs of each hospital could be assessed with respect to its expected costs given the
casemix of patients it treats (Fetter, 1991). The intention is to aggregate patients into a manageable
number of clinically meaningful treatment groups within which one could expect to observe a reasonable
homogeneity of costs.
24
Smith
1. Introduction
Measuring value for money in healthcare: concepts and tools
Figure 4: ONS productivity trends, UK NHS, under various assumptions (year 1999 = 100)
106
104
102
100
98
96
94
92
90
88
86
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Key
Series 1: Quantity productivity index, Paasch drug price index, indirect labour input
measure
Series 2: Quantity productivity index, Paasch drug price index, direct labour input
measure
Series 3: Quantity productivity index, net ingredient cost drug prices, indirect labour
input measure
Series 4: Quantity productivity index, net ingredient cost drug prices, direct labour input
measure
Series 5: Quantity and quality productivity index, Paasch drug price index, indirect labour
input measure (from 1999 only)
Series 6: Quantity and quality productivity index, Paasch drug price index, direct labour
input measure (from 1999 only)
Series 7: Quantity and quality productivity index, net ingredient cost drug prices, indirect
labour input measure (from 1999 only)
Series 8: Quantity and quality productivity index, net ingredient cost drug prices, direct
labour input measure (from 1999 only)
Source: Office for National Statistics, 2006
Table 3 illustrates the importance of DRG risk adjustment by reporting average costs for just a small
number of healthcare resource groups (HRGs) in England. The first point to note is the large variations
in average HRG costs between hospitals, as indicated by the wide interquartile ranges. For example, the
lower quartile value for E12 (acute myocardial infarction without complications) is £775, compared to the
upper quartile figure of £1,718. Such variations offer strong prima facie evidence of large variations in
unit costs between hospitals. This may of course be the result of variations in a number of factors, such
Smith
25
Measuring value for money in healthcare: concepts and tools
1. Introduction
as casemix complexity within HRGs, variations in input prices, differences in accounting practice, data
errors or variations in technical efficiency. Efficiency may, in turn, be affected by considerations such as
local capital constraints, differences in the scale of operations and variations in managerial skills.
Furthermore, the average costs of these procedures vary markedly, ranging from over £32,000 (heart
transplant) to £458 (chest pain aged < 70 and without complications). Clearly, hospitals undertake these
procedures in different proportions, so the need for casemix adjustment is manifest.
Table 3: National average reference costs, selected non-elective inpatient healthcare resource
groups, 2005–2006, England
Average
unit cost
Lower
quartile
Upper
quartile
£
£
£
75
32,113
7,895
47,437
792
4,336
1,572
4,995
Pacemaker implant except for AMI,
heart failure or shock
9,575
3,605
1,540
4,068
Pacemaker implant except for AMI,
heart failure or shock – defibrillator
implant and explant only
977
16,725
13,606
20,737
Acute myocardial infarction with
complications
23,219
1,695
1,097
2,401
Acute myocardial infarction without
complications
63,475
1,169
775
1,718
E15
Percutaneous coronary intervention
24,378
3,401
1,109
3,641
E15DF
Percutaneous coronary intervention
– defibrillator implant and explant
only
81
15,906
14,747
20,214
Heart failure or shock age > 69 or
with complications
52,618
1,694
1,208
2,560
Heart failure or shock age < 70 and
without complications
10,009
1,390
854
1,963
Chest pain age > 69 or with
complications
59,057
603
504
1,123
Chest pain age < 70 and without
complications
95,136
458
409
848
Complex elderly with a cardiac
primary diagnosis
45,347
2,088
1,419
3,018
Count
Code
HRG label
E02
Heart transplant
E07
Pacemaker implant for AMI, heart
failure or shock
E08
E08DF
E11
E12
E18
E19
E35
E36
E99
Source: Department of Health website (www.dh.gov.uk/en/Publicationsandstatistics/Publications/
PublicationsPolicyAndGuidance/DH_062884)
26
Smith
1. Introduction
Measuring value for money in healthcare: concepts and tools
The information in table 3 highlights why risk adjustment is essential when comparing hospital costs. In
England, national reference costs have been used to construct the reference cost index for each hospital
trust. This indicates the ratio of the trust’s actual total costs to its expected total costs, if its costs for each
treatment were at the national average level. Expected costs are calculated by multiplying the number
of cases in each HRG by the associated national unit cost and summing across the trust’s activity.
Amongst acute trusts (excluding specialist hospitals), the index in 2004–2005 ranged from 21 per cent
below expected costs (West Suffolk Hospitals Trust) to 30 per cent above expected costs (Chelsea and
Westminster Healthcare Trust) after adjusting for the higher input prices in London and the southeast.
Numerous other partial indicators of VfM exist. In hospital care, the most widely used is ‘length of
inpatient stay’ associated with an episode. In its narrowest sense, this indicator merely indicates the
use made of a single hospital resource (its beds). However, it is readily measured, and under certain
assumptions acts as a proxy for use of a broader set of inputs such as the associated personnel, and
therefore offers some more general insight into how technically efficient hospital resources are being
used. Again, the interpretation of length of stay requires careful consideration of casemix and mitigating
considerations. Furthermore, its use illustrates the risk of partial indicators, as a reduction in length of
stay might be secured by sacrificing aspects of treatment quality or by shifting costs onto other parties
(such as patients or social care).
None of the measures mentioned above captures variations in quality between hospitals. Do the cost or
process variations reflect to some extent unmeasured variations in health outcomes (such as mortality)?
To the extent that measurement instruments permit, it is therefore desirable to view measures of cost
efficiency and technical efficiency in the light of quality indicators. To date, the emphasis has been on
various measures of mortality associated with hospital care. However, there is currently a concerted
move towards more general collection of outcome data in the form of patient-reported health outcomes
and patient experience surveys (see section 2). To get a full picture of VfM, it is desirable to integrate
these measures into the analysis and move towards more complete measures of cost-effectiveness.
The types of partial VfM indicators routinely used in healthcare are illustrated by the ‘Better care, better
value’ productivity indicators published quarterly by the NHS Institute for Innovation and Improvement for
NHS organisations (summarised in table 4; details at www.productivity.nhs.uk/). These focus on various
aspects of the care process (such as length of stay and use of day case surgery), use of resources
(such as consultant productivity and staff sickness absence) and financial control. It is noteworthy that
many of the indicators associated with patient care are adjusted for the characteristics of the patients
or the underlying population, illustrating the importance of casemix adjustment. Such measures offer
useful insights into the VfM of various aspects of the transformation from inputs into physical outputs of
healthcare. However, it should be noted that there is no attempt to integrate outcomes measures into this
set of indicators.
Smith
27
Measuring value for money in healthcare: concepts and tools
1. Introduction
Table 4: NHS Institute for Innovation and Improvement: ‘Better care, better value’ indicators
Clinical productivity
1. Length of stay
Adjusted for age, sex, diagnosis, method of admission and social
deprivation
2. Day case surgery rates
Percentage of a basket of 25 relevant procedures performed as a day
case
3. Pre-operative bed days
Pre-operative bed days as percentage of all bed days associated with
operations
4. Variation in surgical
thresholds
Number of operations from a basket of five relevant procedures relative
to expected number given the PCT population
5. Variation in emergency
admissions
Actual emergency admissions as ratio of expected level, given the age,
sex and need of the population for 19 conditions
6. Variation in outpatient
appointments
Actual first attended outpatient appointments as ratio of expected
numbers given the age, sex and need of the local population
Finance
7. Financial balance
Whether the organisation is heading for financial balance at the end of
the financial year
8. Cash flow
Actual year to date cash drawings compared to planned year to date
cash drawings
9. Monthly ‘run rate’
Variance between actual surplus/(deficit) for the last month and planned
surplus/(deficit), as a percentage of total planned income
Prescribing
10.Low-cost statin
prescribing
Number of prescription items for low-cost statins as a percentage of
the total number of prescriptions for all statins (excluding combination
products)
Procurement
11.Uptake of national
Uptake of national framework agreements
framework agreements
Workforce
12.Staff turnover
Number of full-time equivalent leavers from an individual organisation as
percentage of average numbers in post
13.Sickness absence
rates
The number of full-time equivalent staff days lost to sickness absence as
percentage of staff in post for the time period
14.Agency costs
Amount spent on agency staff, expressed as a percentage of paybill plus
agency spend
15.Consultant productivity Consultant activity (finished consultant episodes and cost-weighted
activity) in relation to national patterns
Source: www.productivity.nhs.uk
28
Smith
2. What are the components of VfM?
Measuring value for money in healthcare: concepts and tools
2. What are the components of VfM?
Section 1 discussed the variety of VfM concepts that exist, and the challenge of interpreting incomplete
or imperfect VfM measures. Whatever concept of VfM or operational measure is being used, it is
essential to have a clear understanding of the component parts of the VfM calculation, to be able to
understand its strengths and limitations, and to determine priorities for further data collection or analysis.
This section examines the building blocks of any VfM concept in healthcare. Section 2.1 discusses the
notion of the ‘value’ of the various health system outputs, while 2.2 examines how those values might be
quantified. Section 2.3 examines the resource inputs of the health system (the money side of VfM), and
2.4 discusses the important issue of environmental constraints on performance, which may be thought of
as uncontrollable inputs. Section 2 concludes with a discussion of the important issue of whether a short
or long run time horizon is being adopted for the analysis.
2.1. What is valued?
Once the nature of the entity under VfM scrutiny has been established, the first question to ask is what
its valued output is. Some activities and outputs – such as the time spent between patient visits by
community nurses – may be of little value to anyone. What matters is how much society values the
various health sector outputs. In this section, we therefore discuss the outcomes of the health system
that society values, and how these outcomes might be measured.
There is general agreement that the prime objective of healthcare is to improve health. Alongside this,
we also consider three other important categories of objective: responsiveness to patients’ needs,
addressing inequalities, and broader economic objectives. This list of outcomes is not exhaustive. In
many countries, the degree of financial protection from catastrophic expenditure offered by the health
system is an important outcome measure. This insurance role is often taken for granted in debates
in most developed countries. And the health system can also offer substantial benefits to citizens by
reassuring them that, should the need arise, relevant healthcare will be made available. I do not consider
this important reassurance role here. Section 2.1 concludes with a discussion of the more prosaic
measures of healthcare activity and outputs that are often used in VfM studies in the absence of proper
outcome measures.
2.1.1. Health gain7
The most immediate outcome of healthcare is the additional health conferred on the patient, and the
case for defining health system outcomes in terms of health gain is manifest. For most patients and
carers, health gain is the central indicator of the success of an intervention. A focus on these healthrelated outcomes directs attention towards the patient (rather than towards the outputs produced by
the organisation). Moreover, some widely accepted measures of health gain (such as the change in
QUALYs) are independent of the technologies used to deliver care, obviating the need for detailed
scrutiny of the physical actions of organisations when comparing performance.
The measure of health gain should indicate the value added to health as a result of contact with
the health system. Such measures of added value are routinely deployed in other sectors, notably
7 Much of the health services research literature refers to health gains as the ‘outcomes’ of an intervention. This is understandable
since health is the prime focus of most health service interventions. However, I avoid the convention in this paper, as I wish to
reserve the term ‘outcomes’ for all the valued outputs of the health system, which sometimes extend beyond health gain.
Smith
29
Measuring value for money in healthcare: concepts and tools
2. What are the components of VfM?
school education. A central measure of school performance is the contribution made to improving the
educational attainment of pupils. One measure of educational attainment is the exam grades obtained,
which are partly a function of the efforts of the school and partly a reflection of the inherent ability of the
pupil and other external circumstances. Thus, although there are great variations in the abilities of pupils
taught by different schools, the contribution of schools cannot be gleaned solely by reference to their
crude exam results. To make an appropriate comparison, the ability of pupils must be separated from
the school effect. Well-established methods have therefore been developed to measure pupil abilities
at entry to the school, and subsequently to compare exam grades in relation to this baseline, yielding a
measure of educational ‘value-added’ (Goldstein and Spiegelhalter, 1996).
While the concept of value-added is relatively straightforward in the education sector, it has proved more
challenging to make operational in the health sector owing to the much greater heterogeneity of service
users and greater intrinsic measurement difficulties. The fundamental challenge is that, outside a clinical
trials setting, it is rarely possible to observe a baseline: the health status that the patient would have
enjoyed in the absence of an intervention. Although health status measurement is becoming increasingly
routine in many operational healthcare settings, it has mainly involved comparisons of health states
before and after the intervention. This often yields useful information with which to compare different
providers, and is likely to be the basis of most VfM analyses. However, it is worth noting that it cannot
offer a definitive measure of the health gain secured from treatment.
To illustrate the concept of health gain, consider the two diagrams in figure 5. The first, figure 5.1, shows
the progress of health status (measured in QALYs) across a specific individual’s lifetime if they were not
offered healthcare. Total quality-adjusted life years is indicated by the light shaded area. Note the decline
in health status associated with no treatment for this particular individual, which starts in her 40s and
leads to death at about age 70. In figure 5.2 the individual is offered successful treatment at about age
50 that increases both the length and quality of her life, as indicated by the grey shaded area (a). The
size of the area signifies the health gain of treatment, as measured in increased QALYs.
Figure 5: Expected quality-adjusted life years without and with treatment
5.1 Expected QALYs with no treatment
5.2 Expected QALY gain from successful treatment
1.0
Health status
Health status
1.0
0.5
0.0
0
40
Age
QALYs without treatment
80
(a)
0.5
0.0
0
40
80
Age
(a) QALY gain from successful
treatment
An important issue in performance measurement is the valuation of adverse outcomes. Figure 6.1
illustrates the QALY loss that occurs if the same treatment is offered but proves unsuccessful, leading
in this case to rapid deterioration in health and death. Compared with ‘doing nothing’ the patient suffers
a loss of QALYs equal to the light shaded area (b). But compared to a successful outcome, the loss is
30
Smith
2. What are the components of VfM?
Measuring value for money in healthcare: concepts and tools
even greater. The total QALY gain secured by averting an adverse outcome is the sum of (a) plus (b), as
illustrated in figure 6.2.
Figure 6: Expected quality-adjusted life years with unsuccessful treatment
6.1 Expected QALY loss from unsuccessful
treatment
6.2 Expected QALY gain from averting unsuccessful
treatment
1.0
0.5
0.0
Health status
Health status
1.0
(b)
0
40
Age
(b) QALY loss from unsccessful
treatment
80
(a)
0.5
0.0
(b)
0
40
80
Age
(a) QALY gain from successful
treatment
(b) QALY loss from unsccessful
treatment
In practice, of course, much of the information on which diagrams such as this are based is not available.
An important modelling requirement is therefore to infer total health gain from a very small number of
health status measures. Furthermore, any operationally practical measure of health status is likely to be
quite rough and ready, with a risk of reporting bias.
However, a number of well-established measurement instruments have been developed that can be
used to collect before/after measures of treatment effects. These take the form of patient-reported
outcome measures (PROMs) such as the EQ5D and the SF-36 (EuroQol Group, 1991; Ware and
Sherbourne, 1992). There remain many unresolved issues surrounding the precise specification
and analysis of PROMs, and a concern that these generic measures of health status are not
sensitive enough to capture variations in health gain in certain specialties, such as mental health and
ophthalmology. However, for most healthcare, their use is a fundamental requirement for measuring
patients’ health status before and after treatment. Widely accepted methodologies now exist for
translating health status measures into the health gains offered by treatments in the form of QALYs.
In contrast to its highly developed role in assessing individual treatments, the use of health status
measures in the assessment of organisational performance assessment is in its infancy. Historically,
there has been little routine collection of health status measures capable of informing VfM comparisons
across treatments, practitioners and organisations. Recent research has indicated that the cost of
collecting patient-reported outcome measures is low and compliance is high (Smith, Cano et al, 2005).
There is therefore a strong case for making such collection mandatory across all relevant healthcare
as a basis for carrying out comparative performance assessment for use by governments, purchasers
and patients. The NHS has made a start in this respect by requiring, from 2009, collection of EQ5D and
condition-specific measures for four procedures: unilateral hip replacement, unilateral knee replacement,
groin hernia repair, and varicose vein procedures.
Smith
31
Measuring value for money in healthcare: concepts and tools
2. What are the components of VfM?
2.1.2. The patient experience
Quite apart from health gain, patients are becoming increasingly vocal in demanding that healthcare
should be responsive to patient concerns beyond the health effects of treatments. This concern with
the patient experience covers issues as diverse as promptness, autonomy, empowerment, privacy
and choice. Many argue that these concepts should be incorporated into all VfM analyses when they
make a clear contribution to patient well-being. In the UK, one of the biggest concerns in this area has
been various aspects of patient waiting time. However, there is evidence that the UK also scores poorly
on other elements of responsiveness, such as communication between doctor and patient (Blendon,
Schoen et al, 2003).
It is unusual for VfM studies to incorporate considerations related to the patient experience. An important
exception was the WHR 2000, in which WHO developed the concept of the responsiveness of the health
system. This seeks to reflect the extent to which the health system succeeds in being user-oriented
across a number of domains: personal autonomy; choice of providers and treatments; communication
between patients and clinicians; confidentiality; dignity; quality of basic amenities; prompt attention;
and social support. However, although the report contained a useful discussion of the concept of
responsiveness, it was undermined by weak measurement methods. More recent work in the World
health survey has sought to address the issue of responsiveness more satisfactorily (Üstün, Chatterji et
al, 2003). Table 5 gives examples of questions the survey asked in each of the domains.
Table 5: Example questions used to measure responsiveness in the World health survey
Domain
Question
Autonomy
How would you rate your experience of being involved in making decisions
about your healthcare or treatment?
Choice
How would you rate the freedom you had to choose the healthcare providers
that attended to you?
Communication
How would you rate your experience of how clearly healthcare providers
explained things to you?
Confidentiality
How would you rate the way your personal information was kept confidential?
Dignity
How would you rate the way your privacy was respected during physical
examinations and treatment?
Quality of basic
amenities
How would you rate the cleanliness of the rooms inside the facility, including
toilets?
Prompt attention
How would you rate the amount of time you waited before being attended to?
Access to family and How would you rate the ease of having family and friends visit you?
community support
There are, of course, many other ways of conceptualising and measuring the patient experience. For
example, Coulter and Ellins (2006) consider concepts such as patient satisfaction, doctor–patient
communication, psychological well-being, self-efficacy, and patient involvement and empowerment.
Notwithstanding the complexity of such concepts, many survey instruments are now being deployed
routinely to measure the patient experience. These are often extensive in scope, and therefore difficult
to distil into a single number that would be suitable for use in an operational VfM assessment. However,
they capture a great deal of information that could, in theory, be used for informing VfM measures. The
challenge for future research is to find satisfactory ways of condensing the mass of data contained
in surveys into a small number of useful summary measures of responsiveness (Coulter and Magee,
2003).
32
Smith
2. What are the components of VfM?
Measuring value for money in healthcare: concepts and tools
2.1.3. Inequalities
Most health systems pursue equity goals, indicating a concern with variations in access to healthcare
and variations in health itself. These concerns play an important role in public debate about the NHS, but
they are often poorly articulated and not well measured. Furthermore, equity goals sometimes conflict
with other goals such as health maximisation (Williams and Cookson, 2000), and the extent to which
health service organisations should be held accountable for equity issues is contested.
There are two broad schools of thought on how to handle equity issues in VfM analysis. In one sense,
the fairness of the system can be thought of as just another outcome. For example, the WHR 2000
included measures of inequalities in health and inequalities in responsiveness that formed two of the
five outcomes used by WHO as the basis of its measure of system attainment. The challenge using this
approach is that there is no consensus on how to conceptualise or measure inequalities. For example,
should the emphasis be on inequalities between individuals within society or between specific groups
within the population? And, in either case, how should the inequalities be measured? Personal values
and political judgements will necessarily be reflected in any solution to these questions.
Alternatively, some argue that, if equity concerns are important, outcomes (such as health gains)
should be differentially weighted according to who receives them For example, a QALY gained by a
disadvantaged person should be valued more highly than a QALY secured by the rest of the population
(Williams, 1997). Once this principle is agreed, of course, it begs the question of how large the variations
in the valuation of a QALY should be. Work in quantifying the valuation attached to achieving equity
goals is in its infancy (Dolan, Shaw et al, 2005).
2.1.4. Externalities and broader economic outcomes
There is increasing recognition that healthcare yields valued outcomes beyond the immediate health
benefits to the patient. Many treatments offer broader social and economic benefits to patients,
for example in the form of reduced private care costs or improved opportunities to seek out paid
employment. Treatments might also offer analogous benefits to patients’ families in the form of a reduced
need for caring for the patient and increased employment opportunities. An even broader perspective
might extend the notion of value to benefits to society, such as reduced demands on social care
agencies and charities, and macroeconomic benefits such as improved productivity of the workforce
(World Health Organization, 2001).
There is currently little consensus on how to measure the broader societal benefits of healthcare. The
dominant methodology has been the ‘human capital’ approach, which seeks to capture the potential loss
of earnings associated with illness. Human capital is, however, often poorly measured, and indicates
only part of the broader economic benefits associated with health improvement. This approach also
implicitly assigns larger weights to health benefits gained by people with higher earning potential, which
appears to contradict some of the principles of equity adopted in the more traditional valuation of health
benefits.
The decision about whether to adopt a narrow focus on the health gains secured for patients or a
broader focus on societal benefits depends on the perspective of the VfM analysis. If the focus is on
the performance of the NHS in producing health, it might be appropriate to examine only the costs to
the NHS and the associated health benefits. However, it may often be more informative to embrace
the private costs and benefits to patients and their families, and the broader economic perspective. Of
course, many of the benefits (and costs) in this domain are more speculative, difficult to measure and
longer term than the immediate health gain of treatment, and this is the main reason why many economic
evaluations of health technologies do not consider them.
Smith
33
Measuring value for money in healthcare: concepts and tools
2. What are the components of VfM?
2.1.5. Outputs: counting activity and processes
Notwithstanding the clearer thinking now emerging on what the health system is seeking to achieve,
measuring many of the eventual outcomes of healthcare is often very difficult. Most VfM analyses
therefore have to fall back on rather prosaic measures of volume of activity and output, rather than
focusing explicitly on desired outcomes. This is patently inadequate, as it ignores the ultimate objectives:
health gain, improved responsiveness, reduction of disparities, and broader economic contributions.
Indeed, increased activity might at times come at the expense of these more fundamental objectives.
Well-established outcome indicators exist for a limited range of treatments, such as post-operative
mortality rates or infection rates. However, most health services organisations collect limited information
about the health outcomes they produce. More commonly, information is only available about the type of
activities undertaken, for example in the form of the numbers of patients treated, operations undertaken
or outpatients seen. Such quantity measures fail to capture variations in the effectiveness, or quality,
of the healthcare delivered. Yet, despite the growing move towards measuring the outcomes of care,
there is often no alternative to using these crude counts of activity or processes as proxies for the value
produced by healthcare. The generic term for such counts is ‘outputs’.
There are often good reasons why VfM measurement should be based on outputs rather than outcomes.
For example, some health outcomes may take years to be realised, and it is clearly impractical to
wait for them to emerge before attempting to assess performance. It therefore becomes necessary
to rely on measures of activities as proxies for outcome. Measuring activities can also address a
fundamental difficulty of outcome measurement – identifying how much of the variation in outcomes
is directly attributable to the actions of the healthcare organisation. For example, mortality after a
surgical procedure is likely to be influenced by many factors beyond the control of healthcare. In some
circumstances, such considerations can be accommodated by careful use of risk-adjustment methods
(see 3.1). However, there is sometimes no analytically satisfactory way of adjusting for environmental
influences on outcomes, in which case analysing the activities of care instead may offer a more
meaningful insight into organisational performance.
In addition, reliance on counts of activities as a basis for VfM analysis may be unproblematic when there
is good research evidence that an activity (such as an inpatient procedure) on average leads to a known
health improvement. Measuring such activities will give a strong indication of expected health outcomes.
However, when using such measures as the basis for comparing the VfM of healthcare organisations,
it is important to understand that there is an implicit assumption that there is no difference in the
effectiveness with which organisations undertake the activity. Where such differences are suspected,
it becomes imperative to augment activity counts with measures of the quality of outcome in individual
organisations. Ideally, these would indicate health gain, but more readily measured proxies, such as
hospital readmission rates, are often used for such purposes.
Thus, although the use of measures of activity and output are often the only practical option available in
a VfM analysis, it is important to keep in mind the limitations it imposes. In particular, one should beware
of two classes of misinterpretation that commonly result from an absence of outcome information.
• First, all else being equal, organisations that undertake more activities will be rated as more
efficient. But some organisations may have developed care pathways and protocols that
minimise the number of activities required to deliver care to a patient. This may eliminate
unnecessary diagnostic tests, for example, and may be an efficient way of organising care.
However, an activity-based VfM analysis may penalise such organisations.
34
Smith
2. What are the components of VfM?
Measuring value for money in healthcare: concepts and tools
• Second, the effectiveness, or quality, of the healthcare delivered is not captured by a count of
activities. For instance, an activity-based analysis will consider operating theatres that incur the
same costs and undertake the same number of operations to be equivalent, even if patients are
more likely to suffer complications or die if treated in one theatre rather than the other.
In short, any scrutiny of VfM based on activities should acknowledge the limitations it introduces by
ignoring effectiveness considerations.
2.2. Valuing system outputs
Measuring the contribution of healthcare organisations would be difficult enough if those organisations
were seeking to provide a single and relatively homogeneous product. But healthcare organisations
are immensely complex entities, undertaking numerous activities and producing multiple outputs with
different levels of effectiveness. We have already noted the difficulties associated with attaching values
to the various outcomes. The generic measures of health status such as the EQ5D have made some
progress in this respect, but their use has remained fairly circumscribed to date.
The key insight related to valuing health system outputs is to note that all health treatments produce
certain characteristics that are valued by patients. In the health domain, these include longer life,
reduced pain and increased mobility. Under the patient experience, they might include promptness,
dignity and empowerment. It is the aggregation of these qualitative characteristics that gives rise to the
outcomes described above: that is, in the last resort, the valuation of outputs requires an assessment of
their contribution to valued outcomes.
How much the various healthcare outcomes described above are valued are, first and foremost,
personal judgements, and there is evidence to suggest that individual citizens vary greatly in the weight
they place on different healthcare outcomes. Some focus principally on health gain, while others place
great weight on aspects of the patient experience. A conventional economic perspective would suggest
that a proper market in health services would allow patients to express their preferences and thereby
reveal the market price of the different outcomes. However, for many reasons, such markets rarely exist
in health services (Smith, 2000). So, in the absence of market valuations (such as prices), someone
on behalf of society has to decide how much each type of outcome is valued. This is rarely a role for
analysts or researchers – rather, it is the legitimate role of politicians.
Choice of values can nevertheless be informed by evidence from a range of sources, such as ‘stated
preference’ economic studies. These form the basis for the development of many of the QALY
instruments discussed earlier. They can also examine citizens’ willingness to pay for different healthcare
characteristics by asking them to trade off factors such as, for example, waiting time, user charges,
distance travelled and quality of care (Ryan and Farrar, 2000).
An example of a discrete choice experiment designed to solicit patients’ valuations of different
characteristics of their treatment is given by Ryan, Bate et al (2001). Patients at a rheumatology
outpatient clinic were asked to rank a number of different scenarios, described by the six characteristics
listed in table 6. Each scenario comprised a different mix of levels of the six characteristics. Statistical
analysis of the patients’ rankings then allowed the researchers to place a quantitative value on, say, an
improvement in health gain (from no reduction in pain to a small reduction in pain) relative to a reduction
in waiting time (for example, from 20 to 10 minutes).
Smith
35
Measuring value for money in healthcare: concepts and tools
2. What are the components of VfM?
Table 6: Characteristics of patients’ outpatient experience
Attributes
Levels
Health gain
Change in pain between appointments
No reduction, small reduction
Process attributes
The medical staff you see
Junior doctor, specialist nurse
Time in waiting area
Up to 10, 20, 30 minutes
Continuity of contact with same staff
No, yes
Phone-in/advice line service
No, yes
Length of consultation
10, 15, 20, 25 minutes
Source: Adapted from Ryan, Bate et al (2001)
The design and interpretation of such studies requires great care. For example, how should the study
population be chosen, and are the results dependent on the way in which the healthcare setting is
described? Furthermore, the studies cannot dispel the wide variety of preferences that appears to
exist amongst any population. Nevertheless there appears to be great scope for extending this type of
methodology to many situations in which decision makers need to trade off benefits along apparently
incommensurate dimensions of care. As well as informing purchasing decisions, this is an area of central
concern to the development of more satisfactory composite measures of performance, such as the NHS
productivity indices produced by ONS, and is a clear research priority.
In addition to such experiments, the increased choice being offered to patients offers more scope for
statistical analysis of preferences through the actual choices they make. Such ‘revealed preference’
analysis is technically challenging, and little use has been made of it hitherto. However, data sources are
increasingly offering the opportunity to examine how patients implicitly trade off characteristics such as
travel time, clinical quality and waiting times in the real choices they make, and it is likely that more use of
this approach for inferring valuations will become feasible in the near future.
2.3. What are the inputs?
Turning to the money side of VfM, the fundamental concern is to identify inputs and attribute costs to
particular activities. It is relatively straightforward to measure expenditure for a hospital, but it becomes
increasingly difficult to attribute costs to smaller units of observation (department, team, surgeon,
patient). There is often a reliance on arbitrary accounting choices, particularly on the allocation of
overhead costs, which may vary considerably between the units being compared.
Also, many health service outputs are produced by different teams working together. For example, staff
from a variety of hospital specialties contribute to providing care to each patient admitted to hospital.
In such circumstances assessing the relative contribution of each practitioner or team to a specific
output may be challenging. Equally, teams within organisations usually draw on joint resources. For
instance, some staff may work in more than one team, such as when a urologist works partly in general
surgery. It may be difficult to determine accurately what proportion of this shared input is associated
with each team. There is therefore often a tension between seeking out operationally useful VfM costing
information on individual practitioners and teams and using larger aggregations such as hospitals, for
which attribution of inputs and outcomes is less problematic.
36
Smith
2. What are the components of VfM?
Measuring value for money in healthcare: concepts and tools
Moreover, it can be quite challenging to match costs to outputs across time. Some contemporary
costs relate to contemporary outputs, while others relate to future outputs. And conversely, some
contemporary output relies on expenditure made in previous periods. For example, the considerable
sums spent on medical training generate human capital that is expected to yield benefits well into the
future. The issue of matching costs to outputs across time is considered in more detail in section 2.5.
Furthermore, depending on the perspective being adopted, the analysis of treatment costs might in some
circumstances be limited to the purchase price of the treatment, but at other times there might be a need
to assess the broader social costs of treatment (such as transport costs, the treatment’s impact on the
patient’s carer, or the costs imposed on social care agencies). More generally, there remain a number of
important unresolved technical issues in costing methodology (Mogyrosy and Smith, 2005).
In summary, there is continued disagreement on a) the best way to attach monetary value to resource
use, including capital assets, b) the recommended perspective of the study (narrow health service or
broader societal), c) the appropriate measurement and valuation method of informal caregiver time, d)
the measurement and valuation of the costs of lost worker productivity associated with illness, e) the
additional healthcare treatment costs associated with added years of life, and f) the best technique for
allocating support centre costs to operational units.
Finally, inescapable variations in input prices across the units under scrutiny can also be important when
making VfM comparisons. This is most obvious when undertaking international comparison, when it will
usually be necessary to undertake some sort of currency conversion to secure comparability (Busse,
Schreyögg et al, 2008). Input price variations can also be important, even within a country. England
has a long tradition (through the market forces factor) of seeking to adjust NHS expenditure for the very
large variations in the cost of labour and capital across the country, with differences of over 40 per cent
between inner London and some of the rural counties.
In short, although costing methodology has developed considerably over recent years, there are a
number of fundamental choices to be made in enumerating and quantifying the inputs used by the entity
under scrutiny. Although technical, these choices can often have a fundamental impact on the outcome
of any VfM analysis. Indeed, in many settings, proper costing will perhaps prove more of an enduring
challenge than the measurement of outcomes, as some of the challenges of satisfactorily allocating
costs to specific organisations, interventions or patients may prove intractable.
2.4. Environmental constraints
In addition to the challenges in specifying inputs, outputs and outcomes, assessment of VfM in
healthcare is often further complicated by the need to take account of influences on performance that lie
outside organisational control. Numerous classes of factors may influence VfM measures, including:
• differences in the characteristics of citizens being served
• the external environment – for example, geography, culture, and economic conditions
• the activities of other related agencies, both inside and outside the health sector
• the quality of resources being used, including the capital stock
• previous organisational efforts in prevention and health promotion.
Smith
37
Measuring value for money in healthcare: concepts and tools
2. What are the components of VfM?
In the short run, many of these factors may be completely outside the control of the organisations under
scrutiny. They are commonly labelled ‘environmental’ constraints. In particular, citizen characteristics
are often considered to be exogenous influences that determine the context within which the healthcare
organisation must operate, and many of the outcomes it secures are often highly dependent on the
characteristics of the population group it serves. For example, population mortality rates are heavily
dependent on the demographic structure of the population under consideration, and surgical outcomes
usually depend on the severity of a patient’s disease.
There is often considerable debate as to what population factors are considered ‘controllable’ in any VfM
analysis. For example, some critics of the WHR 2000 argued that the HIV/Aids epidemic was a crucial
influence on the poor measured performance of many low-income health systems and had not been
taken into account (Anand, Ammar et al, 2003). Conversely, WHO argued that control of the epidemic
had been amenable to intervention, and performance should therefore be judged without adjustment for
HIV/Aids prevalence rates. In the same way, hospital outcomes may be strongly related to the stage at
which diseases are diagnosed, and there may be debate about the extent to which these are within the
hospital’s control.
Geographical considerations may also play an important part in levels of organisational attainment.
For example, hospital performance may be related to how care is organised in the local community, or
the performance of emergency ambulance services may depend on local geography and settlement
patterns. To the extent that it is feasible, and depending on the purpose of the analysis, it may be
desirable to take account of such variations in any VfM analysis.
The performance of many healthcare organisations is in part dependent on inputs from outside
agencies, such as social care, housing organisations and private families. This too should be recognised
in VfM modelling. For example, many patient outcomes rely on the co-ordinated contributions of
a number of organisations in the form of integrated care. If the performance of only one of these
organisations is under scrutiny, it may be difficult to identify the element of patient outcome that is
attributable specifically to its own endeavours. The danger is either a) its contribution towards integrated
care is ignored in the analysis (under-attribution) or b) the contribution of other external agencies
towards outcome is ignored (over-attribution). Whether these external efforts should be treated as
exogenous may depend on the extent to which the behaviour of external agencies is amenable to
influence by the organisation under scrutiny.
The problem of attribution tends to be exacerbated as the unit of VfM becomes smaller and therefore
more reliant on other parts of the health system in securing desired outcomes. In one sense it is
therefore desirable to adopt the whole-system approach advocated by WHO, under which the outcomes
for the citizen (regardless of which part of the health system secures those outcomes) represent the
outcomes of interest. However, this approach is often unfeasible and may not be helpful because it does
not pinpoint which specific elements of the system are performing well or poorly. Furthermore, it requires
societal values to be placed on a diverse set of outputs.
Notwithstanding the problem of attribution, it is therefore usually more practically helpful to circumscribe
the VfM analysis to clearly defined organisations or practitioners within the health system, even though
this may ignore some of the interactions with other parts of the system. For example, we might be
interested in the cost-effectiveness of individual trauma and orthopaedics specialties in hospitals.
Although this can introduce costing difficulties, taking individuals or teams as the units of analysis has
much to recommend it in comparison with larger organisational aggregations. Their activities may be of a
limited range and can be easily identified and quantified, and the agent accountable for performance can
also be readily identified.
38
Smith
2. What are the components of VfM?
Measuring value for money in healthcare: concepts and tools
One final set of constraints to be noted are those imposed by outside regulators, such as the government
or medical professions, in the form of service standards or training requirements. Such constraints may
affect the extent to which comparable levels of VfM can be achieved by the organisations under scrutiny.
For example, response time standards for ambulances will constrain ambulance station configuration
and affect unit costs in rural areas, and medical training requirements may constrain the extent to which
cost savings through reconfiguration can be achieved. Such constraints might be imposed for reasons
of equity, or to promote broader economic objectives, and may lead to some unavoidable variations in
measures of VfM.
2.5. Short run or long run?
Resolution of some of the debates in VfM discussed in the previous section depends on whether a long
run or short run VfM perspective is being adopted. In the short run, current management must work
largely within current constraints, such as capital configurations and population health characteristics,
and may have little scope to change them. In the longer term, many of these issues are amenable to
purposive improvement on the part of management, which should therefore rightly be held to account for
low levels of associated attainment. In many circumstances it will be useful to undertake both short run
(constrained) and long run (unconstrained) VfM analysis.
The dynamic VfM measurement difficulties can be illustrated in diagrammatic form. The naive (static)
model of VfM in figure 7 assumes that an organisation consumes inputs in the current period and
produces contemporaneous outputs. VfM is assessed by comparing current period outputs (or
outcomes) with current inputs. A typical VfM analysis examines the ratio of outputs to inputs for only a
single time period (say, year t) in this fashion.
Figure 7: The naive (static) representation of value for money
Inputs
year t
Organisation
year t
Valued outputs
year t
Yet the system in year t has usually enjoyed the benefits of past investments. The naive model must
therefore be augmented to incorporate the ‘endowments’ generated for the organisation by its efforts
in previous periods, represented in figure 8 as an additional input to the organisation in year t. And the
organisation also leaves an endowment for future periods in the form of investments undertaken in this
and preceding periods. The endowment might be in the form of real capital (buildings), training of doctors
and other clinicians, medical research, or investment in health promotion and preventive medicine. In
principle, the endowment may be an important aspect of both the inputs to and the outputs of the health
system. In practice, it is very difficult to measure, but it may be a crucial consideration in assessing
longer-term VfM performance.
Smith
39
Measuring value for money in healthcare: concepts and tools
2. What are the components of VfM?
Figure 8: The more realistic (dynamic) representation of value for money
40
Inputs
year t−1
Organisation
year t−1
Valued outputs
year t−1
Inputs
year t
Organisation
year t
Valued outputs
year t
Inputs
year t+1
Organisation
year t+1
Valued outputs
year t+1
Smith
3. Measuring VfM
Measuring value for money in healthcare: concepts and tools
3. Measuring VfM
Having identified the building blocks of VfM, how can they be used to offer meaningful and useful
indicators of performance? In some circumstances, no attempt is made to integrate the various
components of performance discussed above. Instead, as shown in section 1.4, a set of partial indicators
is produced that provides some insights but no definitive judgement. The discussion above suggests that
such indicators may often be helpful, but they can also be misleading. Their incompleteness begs the
question whether the variations observed are really due to differences in VfM or to uncontrollable factors
that have not been captured in the construction of the indicators. As a result, increasing efforts are being
made to develop analytic tools that help to formulate a more rounded judgement of VfM.
More comprehensive approaches have reached an advanced stage of development in the realm of
health technology assessment, often based on randomised clinical trails (Drummond, Sculpher et al,
2005). However, the principles of more comprehensive cost-effectiveness analysis are now also being
tested out in the more challenging arena of performance assessment. This section first summarises
the role of analytic techniques to adjust outputs for variations in environmental constraints (section 3.1).
Even more ambitious are the analytic efforts to embody all relevant inputs, outputs and environmental
constraints into a single productivity model. These are outlined in section 3.2.
3.1. Adjusting for environmental constraints
In contrast with most private sector concerns, public service organisations usually have to operate to
a certain standard, however adverse the circumstances in which they find themselves. In particular,
the NHS is funded by national taxation, and citizens expect a minimum level of service everywhere.
This equity concern gives rise to an important consideration when assessing the VfM of individual
organisations: in assessing performance, to what extent should the external environment be taken into
account, defined, for example, in terms of the population served, local physical infrastructure, other
public agencies or local geography?
In some very special, and highly unlikely, circumstances it may be possible to ignore the environmental
problem when organisations (such as PCTs) have already been fully compensated financially for
environmental circumstances through a well-designed funding formula. A funding formula seeks to
enable organisations to deliver a standard level of service, given environmental factors (Smith, 2007b).
So, if the funding formula is doing its job perfectly, there may be no need to incorporate environmental
factors into the VfM calculations. Instead, all that is needed is to examine the extent to which the required
standards have been secured. In short, one needs to examine only outcomes, and there is no need
to incorporate inputs or environmental constraints into the model because the job has been done in
the funding formula. Of course, in practice, most funding formulae compensate very imperfectly for
environmental factors, so this approach is unfeasible in practice. However, it is important to note the
important link between performance assessment and funding mechanisms.
In practice, the imperfections of funding mechanisms mean that organisations cannot ex ante be
expected to secure identical levels of performance given the resources at their disposal. In whatever way
the uncontrollable environment is defined, some organisations operate in more adverse environments
than others in the sense that external circumstances make achievement of a given level of attainment
more difficult. Therefore, for a given level of inputs, we would expect them to exhibit apparently lower
levels of outputs or outcomes, and apparent VfM. Consequently, a critical requirement of many VfM
analyses is to effect an adjustment for variations in environmental constraints.
Smith
41
Measuring value for money in healthcare: concepts and tools
3. Measuring VfM
Broadly speaking, there are three ways in which environmental factors can be taken into account in any
VfM analysis:
• restrict comparison only to organisations within a similar operating environment
• model the environmental constraints explicitly, as being analogous to other inputs in the
production process
• undertake risk adjustment.
I consider these in turn.
Some of the earliest and most widely deployed approaches towards comparing performance involve
the use of various forms of cluster analysis, which seeks to assign organisations to a small number of
homogenous groups, or clusters (Everitt, Landau et al, 2001). The first requirement of such approaches
is to select various quantitative measures of an organisation’s social, economic and geographical
circumstances relevant to the service in question. Based on these data, a measure of the similarity (or
‘distance’) between each organisation and all others is then calculated. The organisations can then
be clustered into discrete groups exhibiting broadly similar characteristics according to the chosen
measure of similarity. A closely related approach, which is often more satisfactory in the performance
measurement domain, is simply to identify the ‘nearest neighbours’ of each organisation according to
the chosen socio-economic similarity measure, and to use these as the basis for comparison. Such
approaches form the basis for many benchmarking initiatives.
For any observed organisation, cluster analysis (or nearest neighbour analysis) effectively divides the
remaining organisations into just two categories – comparable and not comparable – and comparison
is made only with organisations in the same cluster. Technical choices have to be made regarding
the measure of similarity to be employed and the cut-off criterion for including organisations within a
comparable cluster. The main virtue of clustering techniques is transparency. Although the method
of choosing comparators is technically opaque and vulnerable to arbitrary technical choices, once
the analysis is complete it is a straightforward matter to compare every organisation with the other
organisations within its comparator group. However, the techniques are a very crude method of adjusting
for variations in environment. They assume that organisations are fully comparable with the chosen
comparators but discard potentially useful information about those organisations not selected as
comparators.
ONS produced a cluster analysis of English PCTs based on 42 indicators of social and economic
circumstances in 2001 (National Centre for Health Outcomes Development, 2007). This resulted in 12
discrete clusters, with characterisations such as ‘new and growing towns’ or ‘manufacturing towns’. The
clusters contain relatively homogeneous sets of PCTs which should form a better basis for performance
comparison than, say, regional or national averages. As an example, the cluster defined as ‘industrial
hinterland’ included the following PCTs: North Tyneside; Hartlepool; Knowsley; Darlington; Gateshead;
South Tyneside; Sunderland; Middlesbrough; Tameside and Glossop; County Durham; Sefton; Wirral;
Halton and St Helens; Hull; Stoke On Trent; and Redcar and Cleveland.
The second approach to handling environmental variation is to incorporate environmental factors directly
into the production model, often treating them as exogenous inputs analogous to labour or capital. This
approach effectively generalises the clustering approach by allowing extrapolation from one class of
organisation to another. For example, an environmental factor might be included as an explanatory
variable in statistical models of performance of the sort described further in section 3.2. While leading to
a more general specification of the VfM model than the clustering approach, the direct incorporation of
environmental factors into the statistical model leads to major modelling challenges.
42
Smith
3. Measuring VfM
Measuring value for money in healthcare: concepts and tools
The final method to control for variation in environmental circumstances is the family of techniques
known as ‘risk adjustment’. These methods adjust organisational outcomes for differences in
circumstances before they are deployed in a VfM model, and can offer the most useful balance between
intellectual coherence and practicality when adjusting for environmental factors. Well-understood forms
of risk adjustment include the various types of standardised mortality rates routinely deployed in studies
of population outcomes. These adjust observed mortality rates for the demographic structure of the
population, thereby seeking to account for the higher risk of mortality amongst older people. Likewise,
surgical outcomes might be adjusted for the severity of risk factors, such as age, comorbidities, and
smoking status of the patients treated. The system of diagnosis-related groups (HRGs in England) was
originally developed in order to adjust for the different hospital casemix before making cost comparisons
(Fetter, 1991)
The methods of risk adjustment are often a highly efficient means of controlling an outcome measure
for a multiplicity of environmental factors. Risk-adjusted outcomes can then be entered into the VfM
model without any further need to enter environmental factors. For some health services the methods
of risk adjustment have been developed to a high level of refinement (Iezzoni, 2003). However, it must
be noted that risk adjustment usually has demanding data requirements in the form of information on the
circumstances of individual patients. And even when adequate data do exist, it is often difficult to secure
scientific consensus on the most appropriate way to undertake the risk adjustment. For example, Iezzoni,
Ash et al (1996) examined 14 alternative methods of adjusting hospital death rates from pneumonia for
US hospitals. They found that, although there was some correlation between the results obtained, for
some hospitals, there were serious variations in the performance rankings obtained depending on the
risk adjustment method used.
Whatever the method employed, adjustment for environmental factors seeks to ensure that the VfM
of organisations can be compared on a level playing field. As noted above, this is a highly inexact
science. There is often a balance to be struck between adjusting for every conceivable environmental
reason for variation (in which case every organisation might be deemed unique, and therefore immune
from comparison) and complete disregard of the issue. However, in general, the VfM analysis will lack
credibility if the issue is ignored, so some attention to environmental factors will be required.
3.2. Analytic models of VfM
VfM is, in principle, a simple construct, representing the ratio of outputs, weighted by value, to inputs.
However, the preceding sections suggest that there are numerous analytic difficulties involved in
converting the principle into an operationally satisfactory measure of VfM. Some approaches (such as
the partial indicators discussed in section 1.4) are incomplete and have to be used with caution. The
concern with incompleteness has led to a search for more comprehensive measures of VfM. Yet, as
indicated above, more comprehensive approaches are fraught with difficulties, including the need to:
• specify and measure all outcomes attributable to the units under scrutiny
• attach a weight with which to value all outcomes
• associate all relevant health system inputs to the units under scrutiny
• adjust for environmental factors
• accommodate a longer-term perspective.
Satisfying all these requirements is an immense challenge. Nevertheless, in parallel to the piecemeal
analysis of individual performance measures, a great deal of research effort has also gone into
developing ‘single number’ measures of organisational VfM, under the general banner of ‘productivity
analysis’ (Fried, Lovell et al, 1993; Coelli, Rao et al, 1998). The objective of productivity analysis is to
Smith
43
Measuring value for money in healthcare: concepts and tools
3. Measuring VfM
secure a measure of the cost-effectiveness of an organisation, confusingly referred to almost universally
as a measure of efficiency. Whatever the terminology, the measure of organisational attainment
is defined as a ratio of weighted outputs to organisational inputs, adjusted where necessary for
environmental constraints.
The key contribution of productivity analysis models is a) to adjust for the external environmental
influences on performance and b) to handle the problem of attaching valuations to outputs. Two
approaches have dominated the productivity literature: econometric methods, pre-eminently various
forms of statistical methods such as stochastic frontier analysis (SFA); and the descriptive methods
known as data envelopment analysis (DEA) (Jacobs, Smith et al, 2006). Although these methods
approach the task in radically different ways, they have the common intention of using the observed
behaviour of all organisations to infer the maximum feasible level of attainment and to offer estimates of
the extent to which each individual organisation falls short of that optimum. The methods are technically
challenging, and a full treatment is beyond the scope of this paper. Here I simply seek to give an intuitive
description of each approach.
3.2.1. Statistical methods
Traditional statistical models of healthcare performance usually take the form of a cost function, under
which an organisation’s costs are modelled as a function of a range of organisational outputs.8 The
simplest statistical approach to developing a cost function is to use conventional multivariate regression
analysis, in which costs are modelled as a function of a range of outputs, the organisational environment
and an unexplained error term. This yields an empirical model that predicts an organisation’s expenditure
given its current levels of outputs and environmental circumstances.9 The deviation from this prediction
(the difference between actual and predicted costs) can be used as a basis for estimating the
organisation’s overall efficiency. That is, all unexplained variation from the statistical model is assumed to
be due to inefficiency.
Various refinements of the conventional regression model have been developed to examine
organisational efficiency, including a suite of methods known collectively as stochastic frontier analysis
(SFA) (Kumbhakar and Lovell, 2000). These retain the basic principles of regression analysis, but seek
to decompose unexplained cost variations into random statistical ‘noise’ and inefficiency, the issue of
interest from a VfM perspective. However, SFA requires very restrictive modelling assumptions that are
highly contested and which have led some commentators to question its usefulness (Smith and Street,
2005).
Some of the difficulties brought about by applying statistical methods to a single cross-section of
observations can be obviated by using panel data (that is, a time series of observations for each
organisation, rather than a single measure). The important gain offered by panel data is the vastly
increased ability to distinguish transient (random) variations in performance measures from persistent
(systematic) variations that can form the basis for estimates of inefficiency. However, important technical
assumptions must still be made, for example about how inefficiency is assumed to change over time, and
there is a risk that any model is estimating historical rather than contemporary levels of inefficiency.
8 Some applications have sought to develop the mirror image ‘production function’ of healthcare organisation performance, under
which a single measure of an organisation’s output is modelled as a function of a range of organisational inputs. In general, this
approach poses similar technical challenges but is likely to be less useful from a VFM perspective.
9 It is worth noting that, using the conventional regression model, the coefficient on each explanatory variable in the cost function
offers an estimate of the marginal price of producing the associated output and therefore an estimate of the average implicit
valuation of the output in the sample.
44
Smith
3. Measuring VfM
Measuring value for money in healthcare: concepts and tools
Jacobs, Smith et al (2006) present an application of cost function analysis to 171 acute English hospitals.
They model hospital costs as a function of a range of outputs, including inpatient episodes, outpatient
episodes, accident and emergency attendances, teaching and research, and a number of environmental
factors.10 Using a conventional regression analysis, which treats all of the unexplained variation as
‘inefficiency’, they find that the average level of efficiency is 70.4 per cent. However, when they use SFA,
which treats some of the unexplained variation as random, the average efficiency levels increase to 90.4
per cent.
3.2.2. Descriptive methods
Data envelopment analysis is based on similar economic principles to SFA but uses very different
estimation techniques, based on linear programming models (Thanassoulis, 2001). In summary, it
searches for the organisations that ‘envelop’ all other organisations on the basis of a composite estimate
of VfM. For each organisation, it looks for all other organisations that secure the same, or better, outputs
at the lowest use of inputs. Or, conversely, it can be used to search for the other organisations that use
the same, or lower, inputs to secure the highest level of outputs. For each organisation, the ratio of actual
to optimal performance is referred to as an organisation’s ‘inefficiency’.
Compared to SFA, DEA has some attractive features. It requires none of the restrictive assumptions
required to undertake regression methods. It can handle multiple inputs and multiple outputs
simultaneously, and it requires none of the stringent model testing that is required of statistical
techniques. However, it also suffers from a number of drawbacks. It can be vulnerable to data
errors, because the DEA ‘best practice’ frontier is composed of a small number of highly performing
organisations, and the performance of all other units is judged in relation to that frontier. Therefore, if
the measurement of one key best practice organisation is incorrect, it can result in excessively negative
judgements on many of the inefficient units.
Moreover, from the point of view of ranking organisations, DEA has the profound drawback that it permits
flexibility in the valuation weights attached to each output. The method is agnostic about the valuation
of outputs in the sense that it allows each organisation to be judged using valuations that show it in the
best possible light. Thus each organisation can, in theory, be compared to the frontier according to an
entirely different set of output weights; that is, in the terminology of section 1.2, DEA measures technical
efficiency, not overall cost-effectiveness. In particular, this means that an organisation might be deemed
efficient using DEA, but only if a zero weight is placed on an important output. This appears to contradict
the principle that organisations should be evaluated on a consistent basis, and has also exposed the
technique to fierce criticism (Stone, 2002). For this reason, many commentators advocate the use of
DEA as a useful tool for exploring large and complex datasets but not as regulatory device for passing
judgements or setting VfM targets. Regulators would normally want to apply to all organisations a
consistent set of weights in line with regulatory priorities.
Jacobs, Smith et al (2006) present an application of DEA to 171 acute English hospitals. In the simplest
specification, they use total costs as a measure of inputs, and use inpatient episodes, outpatient
episodes and accident and emergency attendances as outputs. They find that 14 hospitals are 100
per cent efficient and lie on the best practice frontier, and five hospitals have an efficiency level of
less than 50 per cent. The average level of efficiency amongst all hospitals is 74.4 per cent. They then
progressively refine the model to include outputs, such as teaching and research, and include a number
of environmental factors. This leads to a dramatic increase in the number of 100 per cent efficient
hospitals (150 of the 171) and an increase in the average level of efficiency to 98.8 per cent.
10 There was no consideration of the quality of outcomes in this model.
Smith
45
Measuring value for money in healthcare: concepts and tools
3. Measuring VfM
This example demonstrates a number of characteristics of comprehensive VfM measurement,
most notably its sensitivity to the underlying modelling assumptions and the critical importance of
value weights. If more outputs are included, it becomes increasingly difficult to identify best practice
organisations without assigning valuations to the outputs produced. And, other things being equal,
the inclusion of more environmental factors offers organisations more ‘excuses’ for lower levels of
performance. This may be appropriate, but requires careful scrutiny. In practice, any analysis should
examine a range of modelling perspectives in order to identify the sensitivity of judgements to different
technical choices.
46
Smith
4. Conclusions
Measuring value for money in healthcare: concepts and tools
4. Conclusions
In the face of an apparently insatiable demand for treatments, increasing pharmaceutical and manpower
costs and a vibrant market in new medical technologies, the pursuit of VfM in healthcare has become
an imperative for all health systems. Furthermore, the public and its representatives are demanding
increasing levels of accountability for how their tax contributions are spent. Accurate measurement of
VfM has therefore become an urgent priority in healthcare, and this paper has sought to set out the
major issues associated with the pursuit of this objective. The urgency of adopting a VfM perspective
has been heightened further by the recent economic downturn, and the prospect of a highly constrained
financial environment for the foreseeable future. In the same way that pursuit of quality without
consideration of costs may lead to adverse outcomes, the pursuit of spending cuts without consideration
of the health outcomes for patients may also be seriously dysfunctional.
The paper has highlighted two fundamental roles for VfM measurement: prospective assessment of
technologies (for resource allocation purposes) and retrospective assessment of the VfM of individual
providers (performance assessment). In combination, these roles comprise a major element of the
functions of healthcare purchasers (or PCT commissioners as they have become known in England).
The purchasing function is immensely complex and has hitherto been undertaken with only limited
success in most health systems (Figueras, Robinson et al, 2005). Concerted attention to VfM
measurement offers a central focus for improving purchasing for health and healthcare.
The resource allocation role of VfM measurement is relatively well understood, albeit mainly in the
context of individual treatments. Its importance is embodied in current practice of health technology
assessment, and reached its apotheosis in the creation of NICE and analogous agencies worldwide.
Although many contested issues remain, and the practice of introducing VfM considerations into health
technology assessment often introduces unexpected challenges, the methodologies adopted are
relatively well understood and the debates well advanced. Perhaps the major underdeveloped issue
highlighted in this report (section 1.2.1) is how national resource allocation advice can be integrated into
local PCT decision making, where local constraints (and opportunities) become fundamental and there
is a shortage of cost-effectiveness information for many routine treatments. This remains an important
area for further development.
In contrast, the performance assessment role of VfM measurement is underdeveloped. There has
hitherto been a reliance on partial indicators of VfM. These can act as useful diagnostic tools, but they
can also give misleading signals if used carelessly. There is, in my view, an urgent need to complement
these partial measures with more comprehensive measures of VfM performance.
This paper has indicated that the simple view of VfM as the ratio of valued outputs to inputs is
inadequate if one is to move towards useful VfM measures. Figure 9 repeats from an earlier section
the diagrammatic representation of the simple view. This discussion has shown that a more realistic
representation of the VfM measurement challenge would be indicated by something like figure 10. As
well as resource inputs in the current year, we should also consider environmental constraints, policy
constraints and organisational endowments from previous years. On the output side, as well as direct
health improvement outcomes in the current year, we should also consider outputs such as training,
health promotion and other endowments for the future, as well as broader economic benefits from the
health system.
Smith
47
Measuring value for money in healthcare: concepts and tools
4. Conclusions
Figure 9: The naive (static) representation of value for money
Inputs
year t
Organisation
year t
Valued outputs
year t
Figure 10: Towards a more realistic representation of value for money
Exogenous inputs:
other organisations,
population
characteristics, etc
Inputs
year t
System constraints:
policy constraints,
physical constraints,
etc
Endowments
year t−1
Organisation
year t
Endowments
year t+1
Joint outputs:
integrated care,
research, teaching,
etc
Valued outputs
year t
External outputs:
productivity,
independence, etc
In my view, policy makers should work vigorously towards overcoming the current impediments to
establishing a more comprehensive approach to VfM measurement by putting in place appropriate data
collection mechanisms in hard to measure domains, redoubling analytic efforts and commissioning
research to rectify gaps in knowledge. Progress in health technology assessment has shown what can
be achieved. The arguments in favour of pursuing increased comprehensiveness are:
• It offers a rounded assessment of an organisation’s performance across all domains of
endeavour.
• It can facilitate a focus on patient outcomes, regardless of the specific treatments or diseases
under consideration.
• It facilitates communication with ordinary citizens and promotes accountability.
• It indicates which of the entities under scrutiny represent the beacons of best VfM.
• It indicates which entities are the priorities for improvement efforts.
• It can stimulate the search for better data and better analytic efforts across all healthcare.
• It offers mangers of local organisations the freedom to set their own priorities and to seek out
improvements along dimensions of performance where gains are most readily secured, and
does not seek to micromanage (a possible consequence of piecemeal VfM indicators).
48
Smith
4. Conclusions
Measuring value for money in healthcare: concepts and tools
Nevertheless, I acknowledge that the use of comprehensive indicators (in preference to the piecemeal
scrutiny of partial VfM indicators) can in some circumstances be misleading and opaque. The benefits
offered by partial indicators include:
• They can identify serious failings in some parts of the organisation, even if more aggregate
measures of VfM indicate no cause for concern.
• They offer a diagnostic tool for identifying what to attribute poor performance to, and therefore
what remedial action to take.
• They may be the only realistic approach if an attempt to be comprehensive leads to a reliance
on very feeble or opaque data in some dimensions of performance.
• When aggregating different dimensions of performance, comprehensive measures may have to
rely on preference weights that are highly contested.
There will therefore always be a role for partial indicators, especially when the more comprehensive
approach is unfeasible, and also in acting as diagnostic tools.
This paper has indicated that there are many challenges to embedding VfM considerations into the
scrutiny and improvement of the health system. I conclude by summarising what I consider to be the
major tasks for three key constituencies: policy makers and regulators, managers, and researchers.
Policy makers need good VfM for two broad reasons: to help design better (cost-effective) policies, and
to enhance the accountability to citizens of all actors within the health system. In some respects, the UK
has led the world in integrating VfM into the policy process, for example through the creation of NICE
and commissioning the Atkinson review. However, in other respects, progress has stalled. For example,
the Healthcare Commission did not pursue VfM issues with the same vigour as its other responsibilities,
and it is to be hoped that its replacement, the Care Quality Commission, will rectify that lacuna.
It might be felt that measurement of VfM is mainly a technical issue from a regulatory perspective.
However, many measures of VfM are highly dependent on judgements that are properly political rather
than technical (for example, the relative values that should be attached to different outputs). For this
reason, a central requirement is to secure the appropriate involvement of policy makers at all stages of
developing measures of VfM. In enhancing VfM measurement the main roles for policy makers are:
• to develop coherent conceptual frameworks for the design of VfM indicators and their
integration with other performance measures (such as effectiveness indicators)
• to integrate political values into VfM measures when needed
• to mandate the collection of appropriate data
• to assure the independence, quality and comprehensiveness of those data
• to assure high-quality analysis, dissemination and commentary on VfM data
• to promote a more mature national and local debate on VfM
• to integrate VfM into the regulatory process, and act on VfM performance measures in a
proportionate and appropriate fashion.
Managers, including clinical leaders, should seek to integrate VfM into all their decision making.
However, local NHS organisations have traditionally been weak in financial management skills, and
have struggled to make cost-effectiveness central to their business. This is changing, as a result of
the advocacy of the Audit Commission and the NHS Institute and in response to examples such as the
‘service line reporting’ initiative of Monitor. However, there is a large agenda for transferring best practice
Smith
49
Measuring value for money in healthcare: concepts and tools
4. Conclusions
to all NHS organisations. Perhaps the biggest challenge is to ensure that PCTs adopt VfM as the central
pillar of their commissioning role, in deciding what to commission, in determining whether and where to
disinvest, and in checking that providers deliver in line with expectations.
Researchers have made major contributions to the development of NICE cost-effectiveness
methodology. However, they have been less active in the performance assessment domain. This is
manifest in the lack of a coherent intellectual framework underlying many of the VfM performance
assessment initiatives described here. There is therefore considerable scope for further methodological
advance, particularly in developing usable comprehensive performance measures. However,
researchers also have a role in promoting better informed policy and practice, and in nurturing public
understanding of VfM issues. Finally, there remains a widespread failure to integrate VfM considerations
into a great deal of medical and health services research. There is a strong case for greater awareness
of cost-effectiveness issues amongst all health researchers.
The pursuit of VfM is a central concern in all health systems, made strikingly more urgent by the
economic downturn. However, measurement methodology is, and will remain, highly contested and
is at a developmental stage. Yet, notwithstanding its complexity, VfM offers the only concept capable
of offering a unifying framework for assessing all the diverse objectives of health systems. For VfM to
assume the central role that it merits, it is therefore essential that measurement methodologies advance
rapidly, that relevant NHS analytic capacity is put in place at both a national and local level, and that VfM
becomes embedded in all relevant functions of service delivery and policy making.
50
Smith
References
Measuring value for money in healthcare: concepts and tools
References
Anand S, Ammar W et al (2003). ‘Report of the Scientific Peer Review Group on health systems
performance assessment’ in C Murray and D Evans (eds) Health systems performance assessment:
debates, methods and empiricism. Geneva: World Health Organization, pp 839–913.
Atkinson T (2005). Final report: measurement of government output and productivity for the national
accounts. London: Palgrave Macmillan.
Birch S and Gafni A (2002). ‘On being NICE in the UK: guidelines for technology appraisal for the NHS in
England and Wales’. Health Economics, 11, pp 185–191.
Blendon R, Schoen C et al (2003). ‘Common concerns amid diverse systems: health care experiences
In five countries’. Health Affairs, 22 (3), pp 106–121.
Briggs A and Gray A (2000). ‘Using cost effectiveness information’. British Medical Journal, 320, p 246.
Busse R, Schreyögg J et al (2008). ‘Special issue: Variability in health care treatment costs amongst
nine EU countries: results from the HealthBASKET project’. Health Economics.
Castelli A, Dawson D et al (2007). ‘Improving the measurement of health system output growth’. Health
Economics, 16 (10), pp 1091–1107.
Coelli T, Rao D et al (1998). An introduction to efficiency and productivity analysis. Boston: Kluwer
Academic Publishers.
Coulter A and Ellins J (2006). Patient-focused interventions – a review of the evidence. London: The
Health Foundation.
Coulter A and Magee H (2003). The European patient of the future. Maidenhead: Open University Press.
Dawson D, Gravelle H et al (2005). Developing new approaches to measuring NHS outputs and
productivity. Centre for Health Economics Research Paper RP6. York: University of York.
Department of Health and Social Security (1983). Performance indicators: national summary for 1981.
London: DHSS.
Dolan P, Shaw R et al (2005). ‘QALY maximisation and people’s preferences: a methodological review of
the literature’. Health Economics, 14, pp 197–208.
Drummond M, Sculpher M et al (2005). Methods for the economic evaluation of health care programmes
(3rd edn). Oxford: Oxford University Press.
Eddy D (1991). ‘Oregon’s plan: should it be approved?’ Journal of the Amercian Medical Association,
266, pp 2439–2445.
Smith
51
Measuring value for money in healthcare: concepts and tools
References
EuroQol Group (1991). ‘EuroQol: a new facility for the measurement of health-related quality of life’.
Health Policy, 16, pp 199–208.
Everitt B, Landau S et al (2001). Cluster analysis. London: Arnold.
Fetter R (1991). ‘Diagnosis related groups: understanding hospital performance’. Interfaces, 21 (1), pp
6–26.
Figueras J, Robinson R et al eds (2005). Purchasing to improve health systems performance.
Maidenhead: Open University Press.
Fried HO, Lovell CAK et al (1993). The measurement of productive efficiency. Oxford: Oxford University
Press.
Goldstein H and Spiegelhalter DJ (1996). ‘League tables and their limitations: statistical issues in
comparisons of institutional performance’. Journal of the Royal Statistical Society, Series A (Statistics in
Society), 159, pp 385–409.
Gudex C and Kind P (1988). The QALY toolkit, Discussion Paper DP38. York: Centre for Health
Economics, University of York.
Ham C (1998). ‘Retracing the Oregon trail: the experience of rationing and the Oregon Health Plan’.
British Medical Journal, 316 (7149), pp 1965–1969.
Hollingsworth B (2003). ‘Non-parametric and parametric applications measuring efficiency in health
care’. Health Care Management Science, 6 (4), pp 203–218.
House of Commons Health Select Committee (2008). National Institute for Health and Clinical
Excellence. London: The Stationery Office.
Iezzoni LI (2003). Risk adjustment for measuring healthcare outcomes (3rd edn). Baltimore: Health
Administration Press.
Iezzoni LI, Ash AS et al (1996). ‘Judging hospitals by severity-adjusted mortality rates: the influence of
the severity-adjustment method’. American Journal of Public Health, 86, pp 1379–1387.
Jacobs R, Smith P et al (2006). Measuring efficiency in health care: analytic techniques and health
policy. Cambridge: Cambridge University Press.
Kumbhakar SC and Lovell CAK (2000). Stochastic frontier analysis. Cambridge: Cambridge University
Press.
Mitton C and Donaldson C (2004). ‘Doing health care priority setting: principles, practice and
challenges’. Cost Effectiveness and Resource Allocation, 2 (3).
Mogyorosy Z and Smith P (2005). The main methodological issues in costing health care services. A
literature review. CHE research paper 7. York: Centre for Health Economics, University of York.
52
Smith
References
Measuring value for money in healthcare: concepts and tools
National Centre for Health Outcomes Development (2007). Compendium of clinical and health indicators
user guide. London: NCHOD.
Office for National Statistics (2006). Public service productivity: health. London: Office for National
Statistics.
Office for National Statistics (2004). Public service productivity: health. London: Office for National
Statistics.
Peacock S, Ruta D et al (2006). ‘Using economics to set pragmatic and ethical priorities’. British Medical
Journal, 332, p 482.
Ruta D, Mitton C et al (2005). ‘Programme budgeting and marginal analysis: bridging the divide between
doctors and managers’. British Medical Journal, 330, pp 1501–1503.
Ryan M, Bate A et al (2001). ‘Use of discrete choice experiments to elicit preferences’. Quality in Health
Care, 10, pp i55–i60.
Ryan M and Farrar S (2000). ‘Using conjoint analysis to elicit preferences for health care’. British Medical
Journal, 320 (7248), pp 1530–1533.
Smith P (2007a). ‘Performance budgeting in England: experience with Public Service Agreements’ in M
Robinson (ed) Performance budgeting and funding. Washington DC: International Monetary Fund.
Smith P (2007b). Formula funding of public services. London: Routledge.
Smith P (ed) (2000). Reforming markets in health care: an economic perspective. Buckingham: Open
University Press.
Smith P and Street A (2007). ‘The measurement of non-market output in education and health’.
Economic and Labour Market Review, 1 (6), pp 46–52.
Smith P and Street A (2005). ‘Measuring the efficiency of public services: the limits of analysis’. Journal
of the Royal Statistical Society, Series A, 168 (2), pp 401–417.
Smith S, Cano S et al (2005). Patient-reported outcome measures (PROMs) for routine use in treatment
centres: recommendations based on a review of the scientific evidence. London: London School of
Hygiene and Tropical Medicine.
Stone M (2002). ‘How not to measure the efficiency of public services (and how one might)’. Journal of
the Royal Statistical Society, Series A, 165 (3), pp 405–422.
Thanassoulis E (2001). Introduction to the theory and application of data envelopment analysis: a
foundation text with integrated software. Dordrecht: Kluwer Academic Publishers.
Üstün TB, Chatterji S et al (2003). ‘The world health surveys’ in CJL Murray and DB Evans (eds)
Health systems performance assessment: debates, methods and empiricism. Geneva: World Health
Organization.
Smith
53
Measuring value for money in healthcare: concepts and tools
References
Wailoo A, Roberts J et al (2004). ‘Efficiency, equity, and NICE clinical guidelines’. British Medical
Journal, 328, pp 536–537.
Ware J and Sherbourne C (1992). ‘The MOS 36-item Short Form Health Status Survey (SF-36)’.
Medical Care, 30, pp 473–483.
Williams A (1997). ‘Intergenerational equity: an exploration of the ‘fair innings’ argument’. Health
Economics, 6, pp 117–132.
Williams A (1985). ‘Economics of coronary artery bypass grafting’. British Medical Journal, 291, pp
326–329.
Williams A and Cookson R (2000). ‘Equity in health’ in AJ Culyer and J Newhouse (eds) Handbook of
Health Economics. Amsterdam: Elsevier.
World Health Organization (2001). Macroeconomics and health: investing in health for economic
development. Geneva: World Health Organization.
World Health Organization (2000). The world health report 2000: Health systems: improving
performance. Geneva: World Health Organization.
54
Smith